Automatic detection and spatial quantification of trawl-marks in sidescan sonar images through image processing and analysis

M A S T E R T H E S I S

CHARIKLEIA GOURNIA April 2019

T h e s i s supervisor:

Prof. Dr. George PAPATHEODOROU

Environmental sciences School of natural sciences University of Patras Greece

To George Papatheodorou who made this master’s study experience much more beautiful than my fantasy

Supervisor professor:

George Papatheodorou

Examination Committee:

Professor George Papatheodorou

Professor Constantinos Koutsikopoulos

Associate Professor Maria Geraga

Επιβλέπων Καθηγητής:

Γιώργος Παπαθεοδώρου

Τριμελής Εξεταστική Επιτροπή:

Καθηγητής Γιώργος Παπαθεοδώρου

Καθηγητής Κωνσταντίνος Κουτσικόπουλος

Αναπληρώτρια Καθηγήτρια Μαρία Γεραγά

Abstract

Bottom trawl footprints are a prominent environmental impact of deep-sea fishery that was revealed through the evolution of underwater remote sensing technologies.

Image processing techniques have been widely applied in acoustic remote sensing, but accurate trawl-mark (TM) detection is underdeveloped. The paper presents a new algorithm for the automatic detection and spatial quantification of TMs that is implemented on sidescan sonar (SSS) images of a fishing ground from the Gulf of

Patras in the Eastern Mediterranean Sea. This method inspects any structure of the local seafloor in an environmentally adaptive procedure, in order to overcome the predicament of analyzing noisy and complex SSS images of the seafloor. The initial preprocessing stage deals with radiometric inconsistencies. Then, multiplex filters in the spatial domain are performed with multiscale rotated Haar-like features through integral images that locate the TM-like forms and additionally discriminate the textural characteristics of the seafloor. The final TMs are selected according to their geometric and background environment features, and the algorithm successfully produces a set of trawling-ground quantification values that could be established as a baseline measure for the status assessment of a fishing ground.

Keywords: Trawl marks; fishing grounds; automatic detection; side scan sonar; Haar- like features; morphological operations; seafloor characterization

iii Περίληψη

Η αλιεία με μηχανότρατες προκαλεί ένα πλήθος διαταράξεων στον πυθμένα της

θάλασσας, εξέχουσες αυτών τα αλιευτικά ίχνη: γραμμικές ταπεινώσεις του πυθμένα

που δημιουργούνται κατά τη σύρση των αλιευτικών εργαλείων. Η ανάπτυξη της

τηλεπισκόπισης στο θαλάσσιο περιβάλλον φανέρωσε τους σχηματισμούς των

αλιευτικών ιχνών των οποίων η ποσoτικοποίση αποτελεί βασικό μέτρο για την

περιβαλλοντική παρακολούθηση των αλιευτικών πεδίων και την αξιολόγηση

επιπτώσεων. Στόχος αυτής της εργασίας είναι η δημιουργία ενός αλγορίθμου ο

οποίος επεξεργάζεται μεγάλο όγκο καταγραφών του ηχοβολιστικού συστήματος

πλευρικής σάρωσης, εντοπίζει και ποσoτικοποιεί αυτόματα τα αλιευτικά ίχνη. Στην

εργασία παρουσιάζεται επίσης η εφαρμογή του αλγορίθμου σε εικόνες που

συλλέχθηκαν από ένα αλιευτικό πεδίο στον Κόλπο της Πάτρας. Οι εικόνες αρχίκα

αναμορφώνονται με τεχνικές προεπεξεργασίας εικόνων, για να αναλυθούν κατά τον

βέλτιστο τρόπο στο κύριο μέρος του αλγορίθμου. Στο κύριο μέρος της μεθόδου,

μέσω χωρικών φίλτρων εικόνας βασισμένα σε χαρακτηριστικά τύπου Haar,

εξετάζονται όλες οι δομές του πυθμένα σε μια διαδικασία που λαμβάνει υπόψη τους

διαφορετικούς τύπους υφής του πυθμένα. Η πολυεπίπεδη ανάλυση χαρακτηριστικών

στοχεύει στη διάκριση αλιευτικών ιχνών που παρουσιάζονται σε σύνθετα θαλάσσια

περιβάλλοντα. Ανιχνευμένες δομές των οποίων τα φυσικά χαρακτηριστικά

ταιριάζουν με την μορφή των αλιευτικών ιχνών και ανήκουν σε περιοχές με

μορφολογία αντίστοιχη με αυτή που δημιουργούν τα αλιευτικά ίχνη καταμετρώνται

και_ποσοτικοποιούνται.

Λέξεις-κλειδιά: αλιευτικά ίχνη; αλιευτικά πεδία; αυτόματη αναγνώριση;

ηχοβολιστής πλευρικής σάρωσης; χαρακτηριστικά τύπου Haar; μορφολογικές

πράξεις

iv Preamble

Foreword

This master thesis was conducted under the framework of the interdepartmental postgraduate course entitled "Environmental sciences" offered by the School of Natural Sciences of the University of Patras. The purpose of the thesis is to develop an image processing method that will automatically detect and quantify the footprints of trawl-fishing of large- scale areas that are mapped with sidescan sonar imaging system. The output data could be used as scientific database for environmental monitoring and management.

Acknowledgments

There aren't enough words to express how grateful I am that Elias Fakiris made me feel welcome in the laboratory of Marine Geology and Physical

Oceanography of Department of Geology. He shared his knowledge with me and accompanied me till the end of my study, having common pursuit of approaching the logic.

v

Table of Contents

1 Introduction ...... 1 1.1 Monitoring of bottom-trawling impacts ...... 1 1.2 Literature review ...... 7 1.3 Objectives of this study ...... 8 2 Study area and data pre-processing ...... 11 2.1 Study area ...... 11 2.2 Data acquisition and mapping ...... 13 2.3 Data interpretation ...... 15 3 Methodology overview ...... 16 3.1 Preprocessing techniques ...... 18 3.1.1 Intensity normalization ...... 18 3.1.2 Edge preserving smoothing ...... 20 3.2 Seafloor characterization and linear seabed feature detection ...... 22 3.2.1 Multi-scale rotated Haar-like feature...... 22 3.2.2 Seafloor characterization ...... 25 3.2.3 Linear seafloor features detection ...... 27 3.2.4 Morphological operations ...... 29 3.3 TMs extraction and inclusion criteria ...... 30 3.3.1 Geometrical criteria ...... 30 3.3.2 Textural criteria ...... 30 3.4 Trawling-grounds quantification ...... 31 4 Validation data and accuracy assessment ...... 32 5 Results ...... 34 5.1 Trawl-marks extraction and quantification ...... 34 5.2 Accuracy assessment ...... 36 6 Discussion ...... 40 7 Conclusion ...... 45 8 References ...... 48 Appendix...... 63 Publication in journal and abstract for conference paper ...... 63

vi

List of Figures

Figure 1.1 Otter trawl nets and trawl doors ...... 3 Figure 1.2 Satellite bottom trawling imagery ...... 3 Figure 2.1 The study area ...... 13 Figure 2.2 The SSS mosaic ...... 14 Figure 3.1 Overall methodology workflow...... 18 Figure 3.2 Example image of image preprocessing filtering ...... 21 Figure 3.3 Example image of image preprocessing filtering ...... 22 Figure 3.4 Example image of image preprocessing filtering ...... 22 Figure 3.5 The Haar-Like features ...... 24 Figure 3.6 Statistical measures over Haar-Like filter’s responses ...... 25 Figure 3.7 Examples of Anisotropy and Complexity maps ...... 27 Figure 3.8 Example image of Trawl-marks enhancement filtering ...... 28 Figure 3.9 Example image of morphological operations ...... 31 Figure 3.10 Example image of trawl-marks extraction...... 31 Figure 5.1 Automatically and manually extracted trawl-marks ...... 36 Figure 5.2 Trawling-grounds quantification...... 36 Figure 5.3 Accuracy assessment results ...... 38 Figure 5.4 Error estimation ...... 39

vii

List of Tables

Table 1 Results of statistics on the quantification values of TMs length ...... 34 Table 2 Accuracy assessment results ...... 36

viii

1 Introduction

1.1 Monitoring of bottom-trawling impacts...... 1 1.2 Literature review ...... 7 1.3 Objectives of this study ...... 8

“Let the only sound be the overflow, pockets full of stones”.

Francis Eg White & Florence Leontine Mary Welch Musicians

1.1 Monitoring of bottom-trawling impacts

Protecting marine environments is a matter of great scientific and socio-economic concern [1]–[4]. Over the last decades, the progression of remote sensing has brought to light widespread anthropogenic effects on the seafloor [5], [6], caused mainly by fishing, mineral extraction and marine pollution [7]. An emergent theme that has aroused a growing degree of interest is bottom trawling, a method of fishing involving the pull of trawl nets and trawl doors through the seafloor in order to catch bottom-living fish.

Trawl-fishing could lead to benthic habitat degradation with consequent ecosystem

1

alteration [8], [9]. Many gaps are acknowledged in comprehension of the impacts of fishing on the seafloor [10]. These effects are heterogenous and are depended on the spatial and temporal distribution of fishing effort , the habitat type and the environment in which bottom trawl occurs [11].

Direct contact of trawling nets and doors with the seafloor, scraps the sea floor, leaves scars in pairs and causes sediment suspension [12], [13]. Even if it is widely known that bottom trawling causes seafloor trawl marks [14], the penetration depth and the longevity of trawl marks on the seafloor are matters of debate. Some studies have reported trawl marks that are limited to the top centimeters of the seafloor [15] while other mentioned that trawl marks can reach up to 50 cm depth [16]. In the Barents Sea, an area of strong currents, trawl door marks were ceased to be visible within less than five months after the trawl disturbance [17]. Palanques et al. identified the same trawl- marks in side scan sonographs twelve months after a trawling experiment on muddy sediment of an unfinished continental shelf [18]. During the aforementioned experimental trawling study, was also observed a mentionable increase in turbidity of the water column 4–5 days after trawling [18]. The sediment plumes as a consequence of trawling could be observed by satellites along Pecan Island, Louisiana (Figure 1.2).

As for the bottom-trawling effects on the porosity of sediments, the results are incompatible, with some studies reporting no alterations [19] [20] and others reporting a large difference in the compaction of sediments between the trawled and the untrawled sediments [21].

2

Figure 1.1. Otter trawl nets and trawl doors. (Revised from [22]).

Figure 1.2. Satellite bottom trawling imagery from Louisiana. Sediment trails at the ocean surface can be viewed

in the wake of bottom trawling (photo Skytruth)

S.J. de Groot [23] states that as trawl doors pass above the seafloor completely

flatten and eliminate the sand ripples. Bottom trawl has also destructive effects on

3

seagrass beds, seamounts, kelp forests, sponge field and coral reefs that supply shelter, food and nurseries for numerous marine organisms [24] [25]. In some areas, trawling has removed directly 11% of meadow surface [26]. Fossa et al. [27] conducted surveys in Norwegian waters and concluded that the Lophelia pertusa reefs are damaged at about

30-50% by bottom trawling. These biogenic seascapes have the longest recovery rate and is either impossible or very difficult to recolonize [11] [28]. The dramatic effects on emergent epifauna can make greater the predation risk for juvenile fish [10].

Epibenthic animals may be directly killed as bycatch or get damaged and making them more defenseless to scavengers and other predators [29]. Many studies indicate that repeated trawling can transform a community that is dominated by species with relatively large adult size to one dominated by high abundances of small-bodied organisms [11] [30] [31] [32]. Many of the small-bodied species escape probably unscathed from the net and may be able to take advantage of newly exposed food pores that the net is crushing or killing [30] [32]. In some cases, the rest of ocean life is even affected by the removal of a species [10]. Lost static fishing gear (e.g. nets) due to their conflict with seafloor, traps marine organisms and may result in so-called ghost fishing [11] [33]. Commercial fishing is the leading cause of mortality in marine turtles [34] [35] [36] , as trawl nets seem to be the most hazardous in terms of turtle bycatch [37] and consequently is estimated that 8.000 turtles can die every year in the

Mediterranean by bottom trawlers [35].

The changes in oxygen levels [38] and the macrobenthic irrigation activity [39] are the main factors of the changes of geochemical processes in seabed due to bottom trawl. In particular, there is a linkage between bottom trawl and hypoxia at the depth of middle slopes and, consequently, with the restructuring of the benthic megafauna community [40]. Olsgard et al. [41] report that chronic and widespread seafloor

4

disturbance that cause the trawling-induced nutrient flux alterations are capable of significantly inhibiting global nutrient cycles. Analogously, Duplisea et al. [42] estimated that just one trawler passage close to 0.27 km2 h-1 to a depth of ca. 6 cm, can raise the level of nutrients up to 30 times on average in relation to the ambient undisturbed sediment nutrient flux. According to Dounas et al. [43], the sediment disturbance by bottom trawling which increases nitrogen availability and the speed of mineralization process, is followed by higher phytoplankton production if these nutrients have access to the photic zone. However, a full evaluation of the effects of nutrient remobilization requires a careful review with further studies [10]. Moreover, bottom trawling in contaminated sediments could potentially reactivate bioavailable industrial contaminants, release them into the water phase and spread them [44].

It is remarkable that some studies [45] [46] presented geoarchaeological sites that have been heavily trawled, as a result of illegal bottom-trawl. Watling and Norse [47] likened the effects of bottom trawling with terrestrial forest clearcutting, as it transmutes structurally complex habitat which supports high biodiversity, to one flattened, homogenized and with less amount of diversity. Apart from the similarities, Watling and

Norse noted two important factors that differentiate trawling from clearcutting: at first, trawling was mostly unregulated and, moreover, it is practiced over an area 150 times larger than that clear-cut annually.

The amount of the damage done depends on the spatial extent and the frequency of trawling [48]. However, there is a crucial gap in evaluating the trawling impacts originating from insufficient information about the spatial scale and the effort of fishing

[10]. Precise comparison among trawled areas is impractical due to the regional differences in effort reporting methods and records and the disparity in the spatial resolution of the fishing effort data [10]. Following to EU Marine Strategy Framework

5

Directive [49], fishing activities are identified through the data of a GPS-based Vessel

Monitoring System (VMS). However, VMS is not designed for scientific purposes and cannot provide high position quality information due to the low frequency of signal records, the inability to designate whether a vessel is fishing or not and the lack of information about vessel’s gear type [16] [50] [51] [52] [53] [54] [55] [56] [57]. On the other hand, knowledge of TMs’ densities could offer accurate details about the level and the geographic distribution of trawling that are needed for measuring the relevant physical impacts [51],[52].

Almost photo-realistic pictures of large areas of the seafloor are produced in high quality with an acoustic imaging device; side scan sonar [60]. SSS is the most approved system for underwater sound imaging of anthropogenic and natural features of the seafloor over wide-areas [61] [62]. Parts of the system are an underwater transducer

(towfish), a shipboard recording device and their connector cable [63]. Two fan-shaped acoustic pulses are emitted in a sub wide-angle perpendicular to the towfish path, with adjustable frequency, energy and length [64] [65] . These pulses travel in the water zone until they meet objects among their route or the seafloor and then the sound signal is scattered. A proportion of the back-scattered energy arrives back to the transducer and takes the form of electrical signal with the specification of the strength and the travel time [64]. The flow of the electrical signals is recorded as adjoining amplitude traces which synthesize in installments an image of the submarine environment at or near the seafloor [66] . The differentiation of the intensity return values is a function of the porosity and the roughness of the seafloor, the angle in which the acoustic wave hits the seabed surface and the frequency of the system [63]. Objects or features that stand higher or lower of the surrounding seafloor figure shadows in the sonar image. Worthy of attention is also that linear structures are accurately visualized when the towfish fly

6

parallel to their orientation, but may be hidden or not well depicted on tracks vertical to their direction [67]. Trawl door scars are visualized by SSS surveys which are able to ensonify wide areas of the seafloor in relatively short periods of time [68]. Demestre et al. [69] intimated that “the use of SSS records gives an unmistakable location of trawl paths […] and the images of the trawl marks offered an accurate geographic positioning of the fishing activity distribution over the seabed”.

1.2 Literature review

There are numerous paradigms of studies that present SSS images with high density of TMs [12] [14] [17] [18] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79].

The manual interpretation of SSS images is a laborious task that takes considerable time and is important necessity to design computer aided image-processing systems that could automatically detect and quantify TMs. The combination of different seafloor morphologies, the uneven image intensity and the heavy image noise increase the complexity and stand as obstacles in automatic recognition of TMs. González et al. [80] studied an automatic TM detection and classification method on SSS images that was based on the Canny filter edge detector. As TMs take the shape of long, slim, linear scour marks in SSS images [14], in the main stage of their automatic recognition, one may probably emphasize to classic edge detection techniques based on the intensity gradient of the image, such as Canny filter [81]. Classic edge detectors are susceptible to the high variability of the real-world images of deep-sea which contain a large number of edges caused by acoustic radiometric and geometric artifact and therefore these methods could be performed successfully only in selected ideal sub-images, but with low general usability. Tang, Yu and Wang [82] had analyzed the trawling pattern in video images with a neural classifier.

7

Video images are affected by turbidity and can hardly provide a synoptic view of a large fishing ground. Moreover, Malik and Mayer [83], during a multifaceted marine survey identified TMs on SSS records that were not distinctly appeared by video survey. Sams,

Hansen, Thisen, & Stage [84] presented a method for the identification of the seafloor type that is produced by the presence of TMs, but this method did not offer much information for each trawl mark individually. There is an emerging literature of applications concerning seafloor classification using sonar imagery [85] [86] [87].

Texture feature extraction methods are classified in four major categories [88]: (1). statistical (such as gray-level co-occurrence probabilities (GLCMs) [89], (2). geometrical [90], (3). model-based (such as Markov random fields (MRFs,[91]), and (4). signal processing (such as Gabor filters [92], Spatial Domain Filters [93] and Fourier domain filters [90][94]). Spatial domain filters can be much faster, compared to statistical and model-based methods, but they lack in robustness, as they model the environment in a deterministic and rather simplistic manner, incapable of adapting to local conditions. However, they can be used to measure systematic properties of the environment. In principle, they measure the level of contrast between neighbor areas

(usually adjacent ones).

1.3 Objectives of this study

There is obviously an urgent need not just to gather new environmental data but to re-appraise the method of extracting the geographic distribution, the frequency of trawling and their utility for future full evaluation of the effects of bottom trawling.

Simultaneous use of data on characteristics of the seafloor and fishing effort can ameliorate management decisions about how bottom trawling impacts marine

8

environment [10]. To our knowledge, up to today, there is no TM detection algorithm in the literature applied to a large SSS image set that could detect and quantify TMs for environmental impact monitoring use. The purpose of this study is to approach accurately an overview of this issue in both narrow and broad scope terms and propose an advanced algorithm for automatic trawl-marks detection that could be performed for a large trawling ground. This study aims to provide access to unprecedented information about fishing grounds. Two different categories of data are produced in once: estimation of trawl-marks densities and texture characterization. Texture characterization is involved in this study as the concept is not only to concentrate on the linear edge detection but also to define what is and what is not a TM and in which seafloor types it is expected to be detected. This work describes seafloor texture using the contrast textural group, quantified suitably via spatial domain filters (signal/image filtering methods). In the proposed algorithm, spatial-domain filters were employed through

Haar-like features and integral images in a dual-purpose use as they not only produce information about the seafloor type, but can also detect linear seafloor features, like trawl-mark forms. The candidate trawl-marks are undergone morphological operations and geometric analysis that combined with the textural information lead to a comprehensive TM selection.

The automatically detected TMs are quantified in terms of density, length and direction per unit area and the algorithm results are compared to the equivalents of a human interpreter. The SSS dataset used, acquired during a general-purpose research in

Corinth Gulf, Greece, doesn’t meet highest acquisition standards for seafloor target exploration in terms of resolution and acoustic incidence angles, and it thus represents a challenging scenario for the evaluation of the proposed method. The produced data on seafloor characteristics and densities of TMs provide geographic databases for major

9

fishing grounds. The results of the application of this image processing system could be utilized for historical fishing data exploration and extraction, that is precious for monitoring environmental changes related with fishing grounds. A database of TMs densities could offer knowledge on the persistence of trawl marks in different environments.

10

2 Study area and data mapping

2.1 Study area ...... 11 2.2 Data acquisition and mapping ...... 13 2.3 Data interpretation ...... 15

2.1 Study area

The research area is located on the western end of the Gulf of Patras (western Greece,

Mediterranean Sea), almost tangent to the geographic boundary line (Oxeia island to Cape Araxos) which separates the Gulf from Ionian Sea (Figure 2.1). The Gulf of

Patras is a Peliocene-Quaternary basin which is associated with active subduction affected by active faulting of WNW-ESE trend [95]. The Gulf is an interconnected shallow marine embayment between Gulf of Corinth and Ionian Sea [96] and its deepest part lies at the depth of 135 m [95]. Pirros and Glafkos streams are the secondary sediment suppliers of Patraikos Gulf, complementary to the most important sediment provider, Evinos River [96]. Fine sand and silt prevail close to river mouths which exhibit an ongoing fining trend as they move along into deeper water [96].

11

Sedimentation rate ranges levelly from 0,3 to 0,5 mm/yr over the whole Gulf [96]. The seafloor of the research area is made up of Holocene clay and silt with high water content, indicating low energy sedimentation status. Papatheodorou et al. [97] during marine seismic surveys in the Gulf of Patras found the existence of biogenic gases in the sediment pores of the Pleistocene/Holocene interface, forming a gas accumulative horizon. Vertical upward migration of the gases through cracks has created layers of biogenic gas in the upper (Holocaust) layer of sediments with a thickness of 20-30 m, while pockmarks exist on the seafloor and domes that often have a height of 1-7 m.

Coverage of Patras Gulf seafloor with gas charged Quaternary sediments approaches the 70% [98]. Water circulation is mainly caused by wind-driven currents, at least in the winter months and tidal circulation in the Gulf of Patras is characterized by counter- clockwise direction [96]. Noteworthy tidal currents occur principally in the Rion

Narrows [96]. Tidal circulation in the Gulf of Patras is characterized by counter- clockwise direction [96]. Moreover, in the northern part of the bay there is a well- defined upwelling, which does not appear in the southern part [99].

Otter trawling is the most common method of fishing, targeting in soft sediments communities. The study area is estimated to be a bottom trawler fishing ground-hot spot for the 7-8 of the 15 most important commercial species in Greece [100]. Otter trawling is the most common form of fishing, targeting in soft sediments communities. Species richness of interest are: Merluccius merluccius, Micromesistius poutassou, Lophius piscatorius, Mullus barbatus, Mullus surmuletus and Pagellus erythrinus [68]. Due to the physical and mechanical properties of the sediments, TMs are expected to remain as visible linear seafloor disturbance for long period taking also in consideration that the fishing season lasts for 6 months (November-April) every year.

Moreover, the study area considered as an area of important archaeological

12

significance as the crucial Lepanto Naval Battle took place in the area, in 1571, between

Holly League Navy and the Ottoman Nay. Authors of [101] conducted a marine remote sensing survey using a side scan sonar and revealed the presence of numerous small targets, appeared on the sonographs as shallow mounds, over an extensive area. Some targets were considered quite promising and were related to the remnants of the Battle of Lepanto and few others were considered interesting but of ambiguous origin.

Figure 2.1. (a)The study area, (b) The location of Gulf of Patras, (c) Gulf of Patras (revised from [68]).

2.2 Data acquisition and mapping

The survey took place in September 2006, after a closed trawl-fishing season of

6 months. The survey was conducted in a parallel-pass survey design to produce a mosaic with 100% coverage of the seafloor. Six tracks were required with a total length of 13.962 km to cover an area of 2.98 km2. The bathymetry in the research area is gradually increasing from -42 m to the northeast to -53 m in the southwest. Acoustic

13

backscatter data of the seafloor were acquired with an EG&G 272TD sidescan towfish, connected to a top-side processor unit (EdgeTech 4200-P) and the data acquisition software Discover (EdgeTech). The sonar towfish was flown at an average height of 45 m above the seafloor and was operated at a sound pulse frequency of 100 kHz. A

Hemisphere VS101 GPS system was used for the positioning of the research vessel.

The position of the SSS towfish was calculated through the hypotenuse formula taking into account the cable-out and the towfish depth. Due to technical problems, only the right channel of the towfish was in operation and thus the survey-lines were adapted accordingly to achieve the desired overlap (20%) between the corresponding swaths.

The swath range was set to 200 m and the ship speeds ranged between 3 and 4 knots.

Data processing and mosaicking was performed using the "Isis" and "Tritonmap" softwares (Triton Imaging Inc.). Basic radiometric and geometric corrections were applied to the SSS records, including slant-range and Time Varying Gain (TVG) corrections respectively. The produced mosaic (Figure 2.2) had a pixel size of 0.2 m and dimensions of 13.380 x 12.056 pixels.

Figure 2.2. Left: the SSS mosaic of the study area (pixel resolution 0.2 m). Right: (a) area illustrating an abundance of TMs with poor exactness of depiction and (b) area with vivid presence of TMs.

14

2.3 Data interpretation

The interpretation of the 100kHz side scan sonar mosaics revealed three distinct acoustic backscatter patterns in the survey area: (i) uniform low-reflectivity seafloor,

(ii) localized strong backscatter facies in low reflectivity seafloor that well matches with micro-relief seafloor areas, and (iii) numerous, long, linear to curvilinear and parallel to subparallel features on the seafloor (Figure 2.1). The first acoustic backscatter pattern represents fine grained sediments (clay, silt and fine sand). The second one represents high reflectivity mounds. Ground truthing survey showed that those mounds can be considered biogenic mounds (Figure 2.1). These biogenic mounds are quite similar to the coralline algae reefs that have been reported in the Eastern Mediterranean Sea

(Aegean Sea) by the authors of [102]. The linear features of the third acoustic pattern are interpreted as marks produced by bottom-fishing gear (TMs) (Figure 2.1).

15

3 Methodology overview

3 Methodology overview ...... 16 3.1 Preprocessing techniques ...... 18 3.1.1 Intensity normalization ...... 18 3.1.2 Edge preserving smoothing ...... 20 3.2 Seafloor characterization and linear seabed feature detection ...... 22 3.2.1 Multi-scale rotated Haar-like feature ...... 22 3.2.2 Seafloor characterization ...... 25 3.2.3 Linear seafloor features detection ...... 27 3.2.4 Morphological operations ...... 29 3.3 TMs extraction and inclusion criteria ...... 30 3.3.1 Geometric criteria ...... 30 3.3.2 Textural criteria ...... 30 3.4 Trawling-grounds quantification ...... 31

Α multi-stage algorithm was implemented for automated TMs extraction and fishing ground quantification. This method applies spatial domain filtering and morphological operations to achieve an environmentally adaptive TMs detection and quantification, taking into account both TMs morphology and background seafloor type aspects. The procedure begins with subdividing the image into partially overlapping

(sliding) blocks for computational efficiency. Our initial concern was to compensate image noise and amplitude inconsistencies, reform the visual information content of the backscatter images, and then proceed to a more in-depth analysis. Bottom-fishing

16

grounds produce a monadic anisotropic seafloor type described as an otherwise flat seabed form with intense presence of long slightly curved lines [84]. In the main stage of the algorithm, we use Haar-like features that are capable of measuring the contrast between two adjacent areas of the seafloor. We followed work of Fakiris et al. [103] which used spatial domain filtering with Haar-like features for detecting anisotropic

(oriented textures e.g. sand ripples) and complex (e.g. scattered rock - clutter) seafloor areas. Spatial domain filters, such as those based on Haar-like features, provide continuous metrics, which have explicit meanings, correlated to certain—consistent sea bottom characteristics, such as its level of Anisotropy and Complexity under certain scales. Those methods provide filter response maps that can be segmented using simple thresholds into accurate delineations of very specific bottom types. Fakiris et.al [103] used window averaged responses of the Haar-like filter to assess Anisotropy and

Complexity definitions. In this study, effort has been put to designing a new filter with the use of the raw Haar-like filter’s responses in order to scale down the analysis and achieve feature–level segmentation of sub-meter analysis. The responses of the new filter lead to robust TMs detection by producing a map which highlight the locations of the linear seafloor features. The produced response-map is appropriately binarized and morphologically processed, and the TMs are extracted through an inclusion-exclusion procedure accordingly to their geometric, morphological and textural characteristics.

Finally, fishing grounds are quantified in terms of the density, total length and direction of the TMs that they include. The evaluation of the proposed method’s performance was carried out by comparing the automatically extracted TMs with manual. All involved methods were implemented in MATLAB programming tool. The overall methodology workflow is presented in Figure 3.1.

17

Figure 3.1. Overall methodology workflow.

3.1 Preprocessing techniques

3.1.1. Intensity Normalization

The proposed approach embeds a common technique for correcting non-uniform intensity. In underwater acoustic imaging, the absorption of the sound wave energy by the seawater causes negative acceleration to the acoustic echo [104]. The sonar waves also hit the seafloor with wide angular bandwidth [105]. Hence, the returned echo strength that has bounced off the bottom is decreased accordingly to the travelled distance [106]. The fluctuations of the backscattered energy can sensitively vary the levels of side-scan sonar image intensity [107]. Homomorphic filtering, a frequency filtering complementary to an illumination-reflectance model, is applied to control the misbalance of intensity and strength the further processing.

18

Image parts are the reflectance component and the illumination

intensity component, described by a multiplicative relationship:

푓(푥, 푦) = 푖(푥, 푦) 푟(푥, 푦) (1) where 푓(푥, 푦) represents an image, 푖(푥, 푦) is the illumination factor and 푟(푥, 푦) is the reflectance factor [108]. Illumination vary without jerking though the field of view generating low frequencies, unlike the reflectance abruptly changes and is associated with high frequency components [109]. The Homomorphic filtering technique suppresses low-frequency components and is constituted of five steps [110]:

• First, taking the logarithm of equation (1) allows the transformation of the multiplication operation to addition operation: ln(푓(푥, 푦)) = ln(푖(푥, 푦) 푟(푥, 푦)) = ln (푖(푥, 푦)) + ln (푟(푥, 푦))

• The Fourier transform of the log-image is applied to provide the image into frequency domain:

퐹(푊푥, 푊푦) = 퐼(푊푥, 푊푦) + 푅(푊푥, 푊푦)

• Eliminating the low frequencies by the use of a high-pass Gaussian filter H:

퐹(푊푥, 푊푦) = 퐻(푊푥, 푊푦)퐼(푊푥, 푊푦) + 퐻(푊푥, 푊푦)푅(푊푥, 푊푦)

, where

2 2 푊푥 +푊푦 − 2훿2 퐻(푊푥, 푊푦) = (푟퐻 − 푟퐿) (1 − 푊푥 푒 푥 ) + 푟퐿

,푟퐻 and 푟퐿 are the high and the low frequency gain of the filter and 훿푥 is the cutoff frequency.

• Returning to the spatial domain variants of the frequency domain filters by the evaluation of the inverse Fourier-transform.

• Computing the exponential function to obtain the filtered image.

19

3.1.2. Edge preserving smoothing

Considerable emphasis was placed on the most conventional problem of sonar imagery, the injected noise. The acoustic backscatter images are getting acquired and transmitted under the presence of ambient noise [111]. Acoustical noise and internal noise produced by the electrical devices that are used, deteriorate the quality of image

[112]. According to Wilken, Feldens and Wunderlich [113] in most cases side-scan sonar mosaics are affected by visible stripe noise. In our case, during the training process was noticed that both noise and trawl marks are characterized as long lines. The stripe noise and the trawl marks differ in thickness, geometry and intensity. However, if we do not use an effective reduction noise filter, we jeopardize to set stripe noise as candidate line segments which may or may not be portions of trawl marks. Furthermore, the stripe noise engenders impoverishing and fragmentation of the illustrated trawl marks when they intersect. In order to avoid these anomalies, we utilize bilateral filtering with gaussian kernels which offers edge-preserving smoothing. Bilateral filtering was introduced by Tomasi and Manduchi [114] and is a good compromise between complexity and drastic denoising in cluttered imagery. It is a non-linear filter which is synthesized by domain and range filtering. Bilateral filtering run through the image pixel by pixel, replacing every pixel with a weighted

average of the pixels that are close spatially and photometrically with it.

Combined filtering for the point x and a nearby point ξ is given by [115]:

1 퐵 = ∑ 푓(휉 , 푥)푔(퐼 , 퐼 ) 퐼 (2) 푥 퐾(푥) 휉∈훺 휉 푥 휉

훺 is the whole image,

퐼 symbolizes the intensity,

20

1 푑(휉,휒) − ( )2 푓 is the spatial gaussian kernel 푓(휉, 푥) = 푒 2 휎푑

where d(ξ, x) = d(ξ − x) = ||휉 − 푥|| is the Euclidean distance between ξ and x,

1 훿(휉,휒) − ( )2 푔 is an edge stopping function in intensity domain 푔(휉, 푥) = 푒 2 휎훿 where

훿(퐼휉, 퐼푥) = 훿(퐼휉 − 퐼푥) = ||퐼휉 − 퐼푥|| is the distance between the two intensity values

퐼휉 and 퐼푥 ,

and 퐾(푥) is a normalization constant:

퐾(푥) = ∑ 푓(휉 , 푥) 푔(퐼휉, 퐼푥). 휉∈훺

Supposing that points are similar and nearby, the factor 1 takes values close 퐾(푥) to one, allowing the replacement of the point x with an average of the pixels in its

1 vicinity. Inversely, if the pixels are not close, the factor takes values close to zero, 퐾(푥) compelling Bx to bring neuter affects to the centered point x. To conclude, bilateral filtering smooths out the slight difference of neighboring pixels caused by noise, while sharp edges are preserved through the range component.

Figure 3.2. Left, original image with an obvious presence of stripe noise. Right, the result of the pre-

processing stage.

21

Figure 3.3. (a) The original SSS image, (b) the SSS image after application of Homomorphic filter and (c) after edge preserving filtering.

Figure 3.4. (a) The original SSS image, (b) the SSS image after application of Homomorphic filter and (c) after edge preserving filtering.

3.2. Seafloor characterization and linear seabed feature detection

3.2.1. Multi-scale rotated Haar-like features

Haar-like features consist of white and black rectangles whose union is a rectangle. A Haar-like structure is represented as 0s and 1s in a geometrical order and the filter’s response is calculated as the subtraction of the summation of white subregions from the summation of the complementary subregion. Haar-like features have the advantage that they can be calculated directly from an integral image, speeding significantly up data processing. Normal Haar-like features were developed by the authors of [116], their use had spread rapidly but those are not rotation invariant.

22

Various rotated Haar-like features methods have been presented, including that of [103] that covers twelve rotations (Figure 3.5) and is sufficient for depicting features of almost any orientation. The rotated integral image, for a given integer rotation, is calculated by summing the pixels in the relevant aligned quadrant above the given pixel.

Intensity contrasts refer to neighbor areas that have certain spatial relationship

(e.g. certain distance between them). As far as the spatial relationship is concerned, the scale under which the analysis is performed must be specified a-priori. In our case, two- rectangle rotated in twelve angles Haar-like features are able to depict the highlight- shadow scheme of TMs and delineate the various seafloor textures. For a more information-rich result, we implement a multi-scale, rotated Haar-like feature. We utilize a sequence of rotated two-rectangle Haar-like features with common aspect ratio but whose widths respectively correspond on terms of an arithmetic progression. The following function allows to determine the width w and the height h of the sequential

Haar-like features:

s(i) = ( w , h ) = ( z + (i − 1)d , αz + (i − 1)훼d) (3) where i is the ordinal number of each rectangle in the sequential order of Haar-like features, α is the ratio of height to width, z is the width of the first rectangle and d is the common difference between the widths of the rectangles. For this case, the height of the smaller feature is 12 m and its width 0,6 m. The height of the second feature is 16 m and its width 0,8m and the last one has dimensions 20 m and 1 m.

23

Figure 3.5. (a)The sequence of the two-rectangle Haar-like features with different scales and common aspect ratio (b)The Haar-like feature, rotated at twelve angles (revised from [103]).

Next subsections’ filters are performed for each designed scale (s) of the set of

Haar-like features with width (w) and height (h), rotated at direction k, that measure the contrast (c) between two adjacent oblong areas of the seafloor. High values of contrast show that the texture exhibits high differences between local intensities while low ones indicate uniform texture with little or no tonal changes. Local contrast 퐶0 can be expressed as the mean of differences (in grey levels) or as a percentage if normalized by the maximum value that intensities can take. Local contrast normalization for a fixed rotation k and scale s, is defined as is given by [117]:

|퐶0 (푥,푦)| 퐶 (푥, 푦) = 푠,푘 (4) 푠,푘 푅 where R is the intensity range (e.g. 256 for an 8-bit grayscale image) and Cs,k(푥, 푦) ∈

[0,1]. Statistical measures are also constructive for our method such as the maximum, the average (μs,n) and the standard deviation (σs,n) of local contrast along a set of n directions, that are described by Εquations 5, 6 and 7 [117] and an example of their implementation is presented in Figure 3.6:

1 μ (푥, 푦) = ∑푘(푛) 퐶 (푥, 푦) (5) s,푛 푛 푘=푘(1) 푠,푘

1 휎 (푥, 푦) = √ [∑k(n) (C (x, y) − μ (푥, 푦))2] (6) 푠,푛 n k=k(1) s,k s,푛

24

푚푎푥Ck ≥ Ck , ∀ 푘 ∈ [1 , 푛] (7)

Figure 3.6. (a) The original SSS image, (b) and (c) the mean (μ) and the std (σ) values respectively of the filter responses across all directions, (d) the Maximum Filter Responses (MFR), (e) the directions (k) for which the maximum filter responses occur.

3.2.2. Seafloor characterization

Texture can be described according to its two main characteristics: its contrast and orderliness, forming two groups of qualitative terms Image texture, defined as a function of the spatial variation in pixel intensities, is a topic investigated by many researchers in the last few decades [85] [89] [118]. A common application of image texture is the recognition of image regions using texture properties. The textural properties can characterize a seafloor region as rough or smooth, varied or homogeneous, repetitive or random, and hence can help in distinguishing between different areas and features in the images. Haar-like features’ filtering can be used to quantify both contrast and orderliness, as a single feature can assess the local variation of intensities but also highlight certain pre-specified structural geometries.

A square moving average filter is implemented with edge dimensions equal to the

25

height (h) of the Haar-like feature that calculates the local averages of μ (called M) and

σ (called Σ) filter’s responses. Based upon the above local averages of filter responses, two seafloor characterization measures: Anisotropy and Complexity, have been estimated, that are described thoroughly in [103]. The Anisotropy As,h,k is equal to the square of the averaged standard deviation (Σ) to the averaged mean (M) ratio of filter responses over a set of k directions [48]:

훴 As,h,k (x, y) = ( )2 (8) 훭

Anisotropy refers to the level of heterogeneity of filters’ responses given for each direction [119] and therefore high values of anisotropy indicates that at least one direction responses overrides the others. Opposingly, when there are no poles apart in filters’ responses regarding to their direction, anisotropy is low valued. Planar seafloor is characterized by the absence of directional features and is represented by low anisotropy values while a trawled seafloor area gives high values of Anisotropy [103].

The complexity is equal to the square average (μ) to the standard deviation (σ) ratio of the filter responses over the set of k directions [103]:

Μ2 Cs,h,k(x, y) = (9) Σ

The complexity filter highlights the high-contrast areas where seafloor features are anarchical distributed with respect to their direction. Biogenic mounds, bedrocks and scattered boulders synthesize a seafloor area that corresponds to high complexity.

The combination of anisotropy and complexity allows the separation of the area into three seabed types: complex, anisotropic and featureless. The anisotropy and complexity threshold values were set manually by the human interpreter, pre-trained either in a small part of the dataset. Setting a well-suited threshold, complex areas can be easily selected and fully represented as shown in figure 3.7. Anisotropy thresholding creates two zones: above the threshold stands the anisotropic areas and below of it, the

26

benign seabed. Planar seabed and Anisotropic-Complex zones are mutually exclusive.

Figure 3.7. (a) The original SSS image A that exhibits plain sand and TMs, (b) the Anisotropy map of image A, (c) the Anisotropy histogram of image A with vertical orange line indicating the Anisotropy threshold, (d) fishing-grounds segmentation of image A using the Anisotropic threshold. (e) The original SSS image B exhibiting plain sand and biogenic mounds, (f) the Complexity map of image B, (g) the Complexity histogram and the Complexity threshold, (h) biogenic-mounds segmentation using the Complexity threshold.

3.2.3. Linear seafloor feature detection

The second utility of multi-scale rotated Haar-like features concerns linear seafloor features detection (segmentation). As the two-rectangle window slides in twelve directions, any linear edge is effectively depicted. A function over the filter responses aims at enhancing the figure of the TMs. Trawl-marks enhancement filtering

27

(TME) works for each pixel of the image block and is defined as:

푇푀퐸 = (푀푅퐹)2 (10) where MRF (Maximum Response Filter) is the maximum contrast over all directions.

Figure 3.8.b shows a filtering result. Simple thresholding is one of the most important approaches to image segmentation. The most widely used thresholding method is to choose a single threshold value as assignment criteria [120]. Sometimes, the fixed thresholding method breaks down in complex images [121] as the weak edges, features and information are getting ignored [122]. If global threshold fails, it would be more appropriate to calculate an adaptive threshold that varies as it passes through the subregions of the image [123]. In Gaussian adaptive thresholding, the threshold value is a weighted average of the pixel values in a manageable sized neighborhood (x,y) around pixel x. The variable threshold (T) is computed by the convolution of the image with the Gaussian function [107]:

푎 푏 푇(푥, 푦) = ∑푘=−푎 ∑푙=−푏 퐺(푠, 푡)푓(푥 − 푘, 푦 − 푙) (11) 푥2+푦2 1 − and G is the Gaussian function with standard deviation σ: 퐺(푥, 푦) = 푒 2 휏2 . 2휋휏2

Figure 3.8. (a) The original SSS image that depicts TMs of various orientations and contrasts, (b) the TME filter over the preprocessed image, (c) threshold segmentation over the TME filter’s result.

For each scale of the rotated Haar-like features a binary image is produced with the extracted TMs. The binary images are combined with the logical operation “union” that synthesize a final binary image which contains each scale’s extracted TMs.

28

3.2.4. Splitting lines at intersections through morphological operations

All white connected pixels in the binary image are labelled as unique linear foreground objects and their properties are measured, such as their length and their orientation. Some lines intersect and their false orientation could lead to mistaken and incomplete analysis. Therefore, the crossed lines must be split and separated before the orientation of the identified objects is computed. The splitting of the lines is achieved through logical and morphological operations. Line skeletonization produces a thinned center line with respect to the length and the connectivity of the original linear object

[109]. Skeleton shape facilitates the localization of the set of the intersection points.

The points where the lines meet, are dilated and converted to circles in order to overlay the start points of the branches of any possible orientation. The set operation

“difference” follows, and the circle-shaped structuring elements are subtracted by the primary binary image. In this way, the perimeters of the line junctions are throwed away from the original binary image and unlinked section lines are revealed. Figure 3.9 below presents each step of the lines spitting method.

Figure 3.9. (a) The original binary image, (b) the objects’ skeletons, (c) conversion of the line intersection

points to circles and (d) the disjoint line segments.

29

3.3. TMs extraction – Inclusion criteria

Via the feature space we seek to substantiate which line structures correspond to trawl marks. TΜs’ selection is designed based on objects’ geometric signature and the background environment features in the affected area. Following, this is a set of criteria that narrows the detected linear features down to those corresponding to TMs have been designed.

3.3.1. Geometrical criteria

Proud seafloor linear structures are accurately visualized when the towfish fly parallelly to their orientation, but may be subdued or hidden on tracks vertical to it [67]

[124]. Given this we excluded segments that have orientation perpendicular plus-minus

5 degrees from the survey line. This is done because a major part of the SSS image noise is in the across-track direction, caused by towfish movements (pitch and yaw), electric and mechanical interferences. The second discrimination of the lines of interest is built upon their physical size. TMs are appeared as long linear troughs [14]. Lines with length more than 10m are kept for further analysis. Shorter lines are excluded as they may be result of the interaction between high sensitivity of image thresholding and complex marine noise, inferior image resolution or inherently changes of the seafloor backscatter strength.

3.3.2._Textural_criteria

TMs are described as long straight or slightly curved lines, sometimes parallel

30

and equidistant, originated in an anisotropic textured area and seem to be uncorrelated with regions of complex or featureless textural seafloor. The presence of TMs brings anisotropic texture [103]. By implication, areas that are characterized by low

Anisotropy constitutes non-TM regions. Objects in the image whose Anisotropy background values lie under the manually chosen threshold are excluded from analysis.

As well, complex areas are not strongly correlated with regions that contain TMs but their boundaries produce sharp edges in sonar imaging that could be confused with

TMs. Thence complex areas are deleted from the regions of interest. The example given here (Figure 3.10) is based on both geometric and textural criteria.

Figure 3.10. (a) The original SSS image, (b) the detected linear features, (c) the final extracted TMs which meet the inclusive criteria.

3.4. Trawling-grounds quantification

The final binary image is divided into 50x50m blocks and their linear elements are analyzed statistically. The statistical analysis includes extraction of: (i) the density of the detected TMs (counts per area unit), (ii) the total length of the detected TMs per area unit (total length of TMs per area unit) and (iii) the mean per block orientation of the detected TMs. These parameters can be directly correlated to the fishing effort applied to a seafloor area.

31

4 Validation data and accuracy assessment

4 Validation data and accuracy assessment ...... 32

The automatic TMs extraction results were evaluated using a dataset of manually extracted TMs. The results obtained by Patsourakis et al. [68] has preceded this, in which the same data were processed and each individual TM was extracted manually, delineated and mapped on ArcGis software. The vector data of the manually digitized

TMs were undergone spatial analysis to derive their mean density, length and direction per 50x50m square area. These results are used as validation data and are compared to the results of the proposed method to assess its significance. Blocks were classified into four categories according to the TMs’ total length. The first category includes the blocks that do not contain any TMs, the second category includes the blocks with total ΤΜs length less than 200 m, the third category contains the blocks with total length between

200 and 400 m and the last category includes the blocks with total length greater than

400 m. Cross tabulation matrix is used to evaluate the one-to-one correspondence of the classes and to identify a possible relationship between two different classes. The

Cohen's Kappa coefficient evaluates the agreement between the corresponding class

32

limits who each classify the blocks into the four mutually exclusive categories and can range from 0 to +1.

Accuracy of automatic TMs detection method is evaluated by comparing the two data sets that are defined by thresholds of the total TMs length that are used as binary classifiers. If a block contains a number of total TMs length bigger than the threshold is considered as ‘positive’ block else is considered as ‘negative’. Depending on whether respective blocks of manually and automatically derived TMs coincide in their

‘positive’ (positive/positive) or ‘negative’ (negative/negative) characterizations correspond to ‘true positive’ (TP) and ‘true negative’ (TN) sections which indicate an equivalence relation between the data sets. Otherwise, when the blocks are oppositely characterized (positive/negative or negative/positive) correspond to ‘false positive’

(FP) and ‘false negative’ (FN) sections that point out commission and omission errors, respectively. For each class, the number of the TP, TN, FP, FN determine the sensitivity

(or recall), specificity, precision and accuracy measurements that are shown in

Equations (12)-(15):

푇푃 푆푒푛푠푖푡푖푣푖푡푦 = (12) 푇푃+퐹푁

푇푁 푆푝푒푐푖푓푖푐푖푡푦 = (13) 푇푁+퐹푃

푇푃 푃푟푒푐푖푠푖표푛 = (14) 푇푃+퐹푃

푇푃+푇푁 퐴푐푐푢푟푎푐푦 = (15) 푇푃+푇푁+퐹푃+퐹푁

Sensitivity, Specificity, Precision and Accuracy measures can range between 0 and 1, where 1 represents perfect correspondence between the length quantification values of the classified data sets, and 0 indicates failure.

33

5

Results 5.1 Trawl-marks extraction and quantification ...... 34 5.2 Accuracy assessment ...... 36

5.1. Trawl-marks extraction and quantification

The algorithm was applied in sliding blocks of the mosaic to achieve lower computational-cost. The detected TMs are shown in Figure 5.1.a. At the front lines of this study were the attainment of deriving results that can be plotted as density maps that define quantitatively the mechanical pressure on the seafloor, visualize the fishing stress and the preferred axles of towing. In Figure 5.2 vector, TM density maps and rose-plots are presented both for the automatic and manual TMs digitization cases, giving the means for visual comparison and qualitative accuracy assessment. The spatial scale analysis of the detected TMs has a significant impact on interpreting effects of bottom-fishing. The statistical results of the spatial quantification of TMs length are reported in Table 1. The results were calculated per 25 × 10−4 km2 and then converted per km2.

TMs Length TMs Length Descriptive Statistics Automatic method Manual method Min (m/km2) 392 896 Max (m/km2) 86,176 90,493 Mean (m/km2) 25,369 26,226 Total TMs in whole area (m/2.98km2) 53,212 51,960

Table 1. Results of statistics on the quantification values of TMs length. The analysis calculates statistics for the blocks that contain TMs.

34

Figure 5.1. (a) A binary image showing the automatically extracted TMs and (b) an image depicting the

manually digitized TMs (dataset from [68]).

Figure 5.2. (a), (b) Quiver plots visualizing the mean length and orientation of the TMs within each analyzed block for the automatic and the manual TM detection cases respectively, overplayed to maps of the total length of TMs per 10.000 m2 (100x100m block) unit area. (c), (d) Rose diagrams of the corresponding

quiver plots.

35

5.2. Accuracy assessment

The accuracy assessment results of the proposed approach are reported in Table

2. Accuracy exhibited high values, ranging from 77% to 91%, with the highest values

being met in the “0” and “>400” classes as well as in the “TM existence – absence”

case. Precision and Sensitivity were lower at the third and fourth cases of high total TM

lengths, indicating that those length ranges exhibit average agreement to the manually

digitized equivalents. The accuracy assessment results of the first and last classes (“0”

and “>400m”) as well as the very high Accuracy of the “TM existent-absence” binary

case, shows that the proposed method can provide accurate quantitative results for

highly trawled areas, but it can also separate impacted from non-impacted ones. What

is more, the proposed method manages to determine TMs directions with very high

precision, as shown in Figure 5.2.c, d.

Categories Cases TP TN FP FN Sensitivity Specificity Precision Accuracy

0 318 755 19 84 0,79 0,98 0,94 0,91

<200m 655 361 103 57 0,92 0,78 0,86 0,86

200-400m 108 796 95 117 0,38 0,89 0,53 0,77 Classes >400m 104 886 111 75 0,58 0,89 0,48 0,84 Existence- absence 755 318 84 19 0,98 0,79 0,9 0,91

Table 2. The accuracy assessment results. Cases correspond to total TM length per analyzed block.

The cross-tabulation matrix of the four classes (regarding certain ranges of total

per unit area TM length) is shown in Figure 5.3.a. The diagonal in the cross tabulation

has the greatest concentration of scores indicating consistency and stable correlation

between the corresponding classes of the two datasets. The class ‘0’ has a 94%

agreement, conveying that FP rates are almost entirely suppressed at the areas where

TMs do not exist. In classes with higher TM densities there is a good agreement

between the manual and automatic TM densities. The Cohen’s Kappa coefficient is

36

equal to 0.65, indicating a substantial agreement of all class ranges that are used to classify the blocks of the manually and automatically derived TMs. The automatically and the manually extracted TMs have the same spatial distribution when blocks are given 0-1 values (TM absence - presence), where the accuracy value is greater than 0.9.

This implies that fishing grounds, as a bottom-type, can very accurately be detected through the proposed method.

High accuracy values are in the diagonal of the accuracy vs threshold diagram

(Figure 5.3.b), indicating highly accurate results as the total-length threshold increases

(excluding more and more low TM density areas from the analysis). This graph was made to be interpreted as the sensitivity of the method to noise, i.e. to areas with few scattered linear element detections that could potentially be false detection (noise).

However, the fact that very low thresholds (no or limited exclusions) have comparatively better accuracies suggests that even areas with very few TMs are very well segmented and don’t need to be considered even partially false. The agreement between the manual and the automatic TMs assessment is strong (>0.9) up to the threshold of 240 m, after which the Accuracy has a slight decrease, and it is improved after 400 m total TM length per unit area. This threshold range between 240 m and 400 m indicates that there is a factor limiting the accuracy of the method in blocks that are within this range, which is most likely correlated with the northwest area where the worst agreement between manual and automatic TM assessment is met, due to increased deviation between the TMs’ and survey lines’ orientations. Next paragraph is trying to further elaborate on that finding.

37

Figure 5.3. (a) The cross-tabulation matrix for the four-classes scenario and (b) accuracy versus TM length per unit area threshold.

Further elaboration on the reasons that hinder method significance in the present case study, revealed that the direction of the TMs in comparison to the direction of the survey lines plays an important role in TM detection accuracies. This is intuitively clear given that TMs and other proud linear features (sand-ripples) are more difficult to be detected when they get perpendicular to the swath direction, where their contrast and shadow length decreases. Figure 5.2.a confirms the above hypothesis, showing an increased error per unit area as the deviation of the TM orientation from the survey- lines’ equivalent gets higher. The total TM length is slightly underestimated when the orientation deviation is less than 350 but it is severely overestimated (up to 50%) for deviations higher than 350 (Figure 5.4). This finding is opposite to what one should expect, i.e. decreased ability of the system to detect TMs as the direction deviation gets higher, leading to increasing their underestimation. This paradox can be interpreted just in one way: what is expected to bias computer aided detections is also biasing human- interpretation. It is true that there are cases within the present dataset where the human interpreter didn’t include intentionally, or even missed, some minute, controversial linear features that actually were TM parts, while the proposed method did.

38

Figure 5.4. Total TM length error per 100x100m versus the deviation of TM orientation from the survey-line’s equivalent. Y-axis lower limit corresponds to 10% TM length underestimation while the upper one to 50% overestimation.

39

6

Discussion

6 Discussion ...... 40

The proposed algorithm enables the automatic detection of TMs that was previously unattainable due to the limitations of the data conditions and the environmental variable characteristics. The automatic TMs detection algorithm was performed in SSS images with high presence of noise, disfigurements and artifacts. The results reveal that the design of the algorithm overcomes the aforementioned limitations, achieving high accuracy in TMs detection and decreasing the false alarm rates. This image processing method was performed to the whole sonar mosaic providing quantitative estimation of trawling impacts on the seafloor. Returning to the aim of this study posed at the beginning of this paper, it is now possible to state that fishing activity had highly localized impact on the seafloor of Gulf of Patras (Table 1), a biologically significant marine area with historical and cultural relevance. The quantified TMs length values were used as a new indicator of pressure of bottom trawling.

The preprocessing stage was a necessary step that successfully normalize the non-uniform intensity, suppresses the noise and at the same time gives at TMs a more discrete and intact appearance in SSS image data. In reviewing the literature, many prior studies in this field [66], [125] have noted the importance of using an uniform

40

backscatter intensity mosaic. Radiometric and geometric corrections is an issue of high importance in acoustic imaging, but in SSS data, where there is no auxiliary information about the exact geometry of the ensonification, their elaboration is difficult [104]–

[107]. The proposed approach embeds a common technique in satellite and medical image processing for correcting non-uniform intensity, the Homomorphic filtering, which reverses the situation of hidden environment in images. Homomorphic filtering helps in TMs separation when the background backscatter intensities change in space, depending on the sediments properties, making TMs in low backscatter intensity soft sediment hardly separated.

The implemented algorithms exhibit low computational costs, utilizing calculations on integral-image, making it suitable for quasi real-time operations. Digital filters based on multiscale, rotated Haar-like features, proved to be capable of detecting

TMs of great variability in curvature, orientations, intensities and contrasts. The smaller rotated Haar-like features can depict more precisely the step edges of a narrow line and fit better in curves of considerable torsion. The longer Haar-like features are able to highlight ramp edges and surpass small gaps of long linear edges. The binarization of the highlighted linear features and the division of the crossed lines was two of the key problems in this method. The adaptive thresholding work effectively and consistently in every region of the image, giving back a detailed binary representation of the original image. Thus, even older TMs, with lower backscattered contrasts are performed as foreground lines. The lines splitting method was achieved in an accurate way by removing the nodes of the skeleton which satisfy high geometrical fidelity to the lines’ representation. The splitting of the lines allows the proper lines orientation measurement that is precious information for a set of operations in this method, such as the TMs quantification in terms of orientation.

41

Thresholding is one of the most important approaches in image segmentation.

The most widely used thresholding method is to choose a single threshold value as assignment criteria [120]. Sometimes, the fixed thresholding method breaks down in complex images [121] as the weak edges, features and information are getting ignored

[122]. To maximize the generalizability of the designed algorithm, we tried to make use as much as possible of adaptive thresholds [123], that require no human decision making to be involved in the process. However, our method still concerns a range of parameters that need to be defined manually. Some of them were set to their most bibliographically common ones (intensity normalization), offering adequate results.

Others were determined using the well-defined geometries of the target seafloor feature

(TMs), such as the Haar-like features, whose dimensions were adapted to common TM geometries met in sonar images. Still, very few parameters needed to be set in a heuristic. The latter are the Anisotropy and Complexity thresholds used for environmental adaptation through seafloor characterization. These thresholds are particularly sensitive to the ensonification geometries (incidence angles, towing-height above the ground) and the equipment used (transmit frequency) that control the shadow extents and the texture contrast of the TMs. However, those are simple thresholds and so they need no time-consuming training phases but rather a couple of sample areas to tune them suitably.

The SSS mosaic encompasses a combination of anisotropic areas with the intense presence of TMs and areas of featureless homogeneous soft sediments. The SSS mosaic also depicts complex areas that represent scattered biogenic mounds. A growing body of literature pays particular attention to the extent and the acoustic pattern of coralligenous formations [102], [118], [126], [127]. Despite the fact that the EU

1967/2006 Mediterranean fisheries management Regulation protects the biogenic

42

habitats, they are commonly impacted by fishing activities [102], [128]. The presence of biogenic mounds creates areas of complex surficial geology which are accurately delineated by the Complexity filter. One unanticipated finding of this study was that there is no linkage of the trawling activity with the biogenic mounds according to the textural analysis and the data interpretation of the mosaic. This is most probably explained by the fishermen's knowledge of this rigorous morphology of the seafloor and the consequent avoidance of surpassing them in order to protect their gears [68].

This is also reaffirmed by the turn of the trawl tracks on the northern boundary of the research area [68] (Figure 11a,b). TMs presence on seafloor implies tonal differences of adjacent image regions creating an anisotropic textured area [103] and seem to be uncorrelated with regions of featureless textural seafloor which is characterized by homogeneity. By implication, areas that are characterized by low Anisotropy constitutes non-TM regions. The implementation of Anisotropy filter facilitates the separation of the two different textures which consequently eliminate the number of false alarms at the zones where TMs are absent, as reaffirmed by the high precision score of the class ‘0’. TMs are described as long straight or slightly curved lines, sometimes parallel and equidistant. Boundaries of biogenic mounds produce sharp edges in sonar imaging that could be confused with TMs and therefore are excluded.

Collectively, the information about the texture of the seafloor has a valuable impact on eliminating false alarms while no classifiers are needed for the seafloor characterization.

Seafloor linear structures are accurately visualized when the towfish fly parallel to their orientation, but may be not clearly depicted or hidden when the towfish track is perpendicular to their orientation [67], [124]. In addition, a major part of the SSS image noise is in the across-track direction, caused by towfish movements (roll, pitch and

43

yaw), electric and mechanical interferences. These factors contribute to the exclusion of the detected line segments that have orientation perpendicular ± 5 degrees from the survey line. Contrary to expectations, total 100 degrees orientation exclusion is an appropriate way to achieve the right balance between a high level of noise-free mosaic and high accuracy in the TMs density estimation. Moreover, as TMs are appeared as long linear troughs [29] in SSS data, short lines are not selected as TMs. Short line segments may be a result of the interaction between high sensitivity of the adaptive thresholding and complex marine noise, inferior image resolution or inherently changes of the seafloor backscatter strength. The data used in the present work was not acquired for the assessment of trawling activity, which would require higher acoustic incidence angles (proximity of the SSS towfish to the ground) but rather for a general-purpose, low-cost and time-limited seafloor research. The swath range was kept high (200 m) and as a consequence the acoustic signature of TMs has low contrast and under- developed acoustic shadows, hindering the quality of the produced mosaic. The direction of the survey lines was set constant, without any multi-angle view compensation. This way, TMs that run parallel to the survey tracks had a more vivid and clear depiction than the ones which stand at an increased angle [29,90], affecting the detection accuracy. Consequently, the design of this survey lacked some sidescan tracks, especially perpendicular to the tracks made, which would offer focused and safer

TMs depiction by viewing them from different angles.

44

7

Conclusion

7 Conclusion ...... 45

Trawling has significant impacts on seafloor habitat, depending on the spatial distribution and the frequency of the fishing effort so that urgent management actions are required. In this paper a new method was presented that achieves automatic detection and spatial quantification of trawl-marks under challenging conditions of acoustic imaging when general-purpose SSS data are used. The algorithm was demonstrated on sidescan sonar images of board-scale and structural complex habitats and successfully detected trawl-marks of great variability with respect to orientation, intensity and contrast. This is the largest study so far documenting a delayed onset of automatic TMs quantification in terms of length and orientation, that is also characterized by high accuracy values. The integration of this image processing method in acoustic remote sensing, provides valuable information about the bottom fishing effort intensity and distribution. The approaches developed prove that there is definitely room for improvement in identifying fishing behavior of trawlers and in evaluating fishing impacts on the seafloor through SSS imaging. Spatiotemporal analyses of the exported data are expected to be useful for the prognosis of the ecosystem’s recovery

45

trajectories in terms of trawling impacts. The proposed methodology is expected to be a useful tool for the protection of the habitat of priority and specifically for Posidonia oceanica meadows and coralligenous formations/ maerl populations.

A natural progression of this study is to implement some of its features in conjunction with machine learning principles through the use of extended sonar datasets of TMs. As in the case of other geological seafloor features (e.g., faults, land-slides, pock-marks etc.) though, open swath sonar datasets are still quite limited to meet the need for extensive training datasets that deep learning or other machine learning methods demand. A major factor for this is that sonars produce multi-aspect, multi- frequency views of the same seafloor target, depending on the acquisition parameters and the type of equipment used, and so a huge number of samples may be needed to adequately include all the possible versions of the same feature in the training dataset.

In addition, backscatter intensity normalization is still a matter of negotiation among acousticians and so datasets between different users are not yet directly comparable.

Although transfer learning is an intriguing option in recent deep learning applications, that could minimize the training-sets needed or even take advantage of extensive datasets in similar terrestrial applications (e.g., radarsat ), it still needs to be extensively validated against real sonar datasets before replacing traditional spatial domain filtering for seafloor feature recognition, as those used operationally in mine counter- measurement tasks [91].

46

8 References

[1] R. M. P. Rosales, “SEAT: Measuring socio-economic benefits of marine

protected areas,” Mar. Policy, vol. 92, pp. 120–130, 2018.

[2] J. N. Sanchirico, K. a. Cochran, and P. M. Emerson, “Marine Protected Areas:

Economic and Social Implications,” in Resources For the Future. Discussion

Paper, Washington, DC, USA, 2002.

[3] N. J. Bennett, “Marine Social Science for the Peopled Seas,” Coast. Manag.,

vol. 47, no. 2, pp. 244–252, 2019.

[4] D. Russi et al., “Socio-Economic Benefits of the EU Marine Protected Areas,”

Inst. Eur. Environ. Policy, London, UK, 2016.

[5] F. K. J. Oberle, P. Puig, and J. Martín, “Fishing Activities,” in Submarine

Geomorphology, A. Micallef, S. Krastel, and A. Savini, Eds. Cham: Springer

International Publishing, Cham, Switzerland, 2018, pp. 503–534. ISBN 978-3-

319-57852-1.

[6] E. Ramirez-Llodra et al., “Man and the Last Great Wilderness: Human Impact

on the Deep Sea,” PLoS One, vol. 6, no. 8, pp. 1–25, 2011.

[7] L. Woodall, C. Stewart, and A. Rogers, “Function of the High Seas and

Anthropogenic Impacts Science Update 2012-2017,” University of Oxford:

Oxford, UK, 2017.

[8] L. Merillet et al., “Using underwater video to assess megabenthic community

vulnerability to trawling in the Grande Vasière (Bay of Biscay),” Environ.

47

Conserv.,vol. 45, pp. 163–172, 2017.

[9] B. Riemann and E. Hoffmann, “Ecological consequences of dredging and

bottom trawling in the Limfjord, Denmark,” Mar. Ecol. Ser., vol. 69, pp. 171–

178, 1991.

[10] Board, O.S, and National Research Council, “Effects of trawling and dredging

on seafloor habitat,” National Academies Press, Washington, D.C.,USA, 2002.

[11] M. J. Kaiser, J. S. Collie, S. J. Hall, S. Jennings, and I. R. Poiner, “Impacts of

fishing gear on marine benthic habitats,” in Conference on Responsible

Fisheries in the Marine Ecosystem, 2001.

[12] M. Robelly and R. J. Beaman, “Spatial analysis of seabed trawl marks,

Frederick Patches, northern Great Barrier Reef,” in SEES Research Student

Conference, Townsville, Australia, 1 November 2011.

[13] J. B. Jones, “Environmental impact of trawling on the seabed: A review,” New

Zeal. J. Mar. Freshw. Res., vol. 26, no. 1, pp. 59–67, 1992.

[14] A. M. Friedlander, G. W. Boehlert, M. E. Field, J. E. Mason, J. V. Gardner, and

P. Dartnell, “Sidescan-sonar mapping of benthic trawl marks on the shelf and

slope off Eureka, California,” Fish. Bull., vol. 97, pp. 786–801, 1999.

[15] M. Brylinsky, A. Gibson, and D. C. Gordon Jr, “Impacts of Flounder Trawls on

the Intertidal Habitat and Community of the Minas Basin, Bay of Fundy.,”

Can. J. Fish. Aquat. Sci., vol. 51, pp. 650–661, 2011.

[16] P. Buhl-Mortensen and L. Buhl-Mortensen, “Impacts of Bottom Trawling and

Litter on the Seabed in Norwegian Waters,” Front. Mar. Sci., vol. 5, no. 42,

2018.

[17] O. B. Humborstad, L. Nøttestad, S. Løkkeborg, and H. T. Rapp, “RoxAnn

bottom classification system, sidescan sonar and video-sledge: Spatial

48

resolution and their use in assessing trawling impacts,” ICES J. Mar. Sci., vol.

61, pp. 53–63, 2004.

[18] A. Palanques, J. Guillén, and P. Puig, “Impact of bottom trawling on water

turbidity and muddy sediment of an unfished continental shelf,” Limnol.

Oceanogr., vol. 46, no. 5, pp. 1100–1110, 2001.

[19] P. Percival, “Impacts of trawl fisheries on marine benthic biogeochemistry,”

University of Newcastle, 2004.

[20] A. W. Simpson, “An Investigation of the Cumulative Impacts of Shrimp

Trawling on Mud Bottom Fishing Grounds in the Gulf of Maine : Effects on

Habitat and Macrofaunal Community Structure,” The University of Maine,

2003.

[21] A. Pusceddu, S. Bianchelli, J. Martín, P. Puig, A. Palanques, and P. Masqué,

“Chronic and intensive bottom trawling impairs deep-sea biodiversity and

ecosystem functioning,” Proc. Natl. Acad. Sci. U. S. A., vol. 111, no. 24, pp.

8861–8866, 2014.

[22] F. K. J. Oberle, P. W. Swarzenski, C. M. Reddy, R. K. Nelson, B. Baasch, and

T. J. J. Hanebuth, “Deciphering the lithological consequences of bottom

trawling to sedimentary habitats on the shelf,” J. Mar. Syst., vol. 159, pp. 120–

131, 2015.

[23] S. J. de Groot, “The Imapct of bottom trawling on benthic fauna of the North

Sea,” Ocean Manag., vol. 9, pp. 213–226, 1984.

[24] C. Wilkinson, Status of coral reefs of the world:2008. Townsville, Australia:

Global Coral Reef Monitoring Network and Reef and Rainforest Research

Centre, 2008.

[25] M. Bharathiamma and U. Bhagirathan, “Impact of Bottom Trawling on Benthic

49

Communities :,” Fish. Technol., vol. 45, no. 1, pp. 1–22, 2008.

[26] S. Kiparissis et al., “Illegal trawling and induced invasive algal spread as

collaborative factors in a Posidonia oceanica meadow degradation,” Biol.

Invasions, vol. 13, no. 3, pp. 669–678, 2011.

[27] J. H. Fosså, P. B. Mortensen, and D. M. Furevik, “The deep-water coral

Lophelia pertusa in Norwegian waters : distribution and fishery impacts,”

Hydrobiologia, no. 471, pp. 1–12, 2002.

[28] M. Gianni, “High Seas Bottom Trawl Fisheries and their Impacts on the

Biodiversity of Vulnerable Deep-Sea Ecosystems : Options for International

Action,” IUCN, Gland, Switzerland, 2004.

[29] C. L. J. Frid, R. A. Clark, and J. A. Hall, “Long-term changes in the benthos on

a heavily fished ground off the NE coast of England,” Mar. Ecol. Prog. Ser.,

vol. 188, pp. 13–20, 1999.

[30] R. Engel, Jonna Kvitek, “Effects of Otter Trawling on a Benthic Community in

Monterey Bay National Marine Sanctuary,” vol. 12, no. 6, pp. 1204–1214,

1998.

[31] S. Jennings, T. Dinmore, D. Duplisea, K. J. Warr, and J. E. Lancaster,

“Trawling disturbance can modify benthic production,” J. Anim. Ecol., no. 70,

pp. 459–475, 2001.

[32] M. Demestre, P. Sánchez, and M. Kaiser, “The behavioural response of benthic

scavengers to otter-trawling disturbance in the Mediterranean,” in Effects of

Fishing on Non-target Species and Habitats, pp. 121–129.

[33] F. Baeta, M. J. Costa, and H. Cabral, “Trammel nets ’ ghost fishing off the

Portuguese central coast,” vol. 98, pp. 33–39, 2009.

[34] L. Alessandro and S. Antonello, “An overview of loggerhead sea turtle (

50

Caretta caretta ) bycatch and technical mitigation measures in the

Mediterranean Sea An overview of loggerhead sea turtle ( Caretta caretta )

bycatch and technical mitigation measures in the Mediterranean Sea,” Rev.

Fish Biol. Fish., vol. 20, pp. 141–161, 2010.

[35] P. Casale, “Sea turtle by-catch in the Mediterranean,” Fish., vol. 12, pp. 299–

316, 2011.

[36] B. P. Wallace et al., “Global patterns of marine turtle bycatch,” Conserv. Lett.,

vol. 3, pp. 1–12, 2010.

[37] A. Lucchetti, C. Vasapollo, and M. Virgili, “An interview-based approach to

assess sea turtle bycatch in Italian waters,” PeerJ, 2017.

[38] J. S. Collie, S. J. Hall, M. J. Kaiser, and I. A. N. R. Poiner, “A quantitative

analysis of ® shing impacts on shelf-sea benthos,” 2000.

[39] M. Trimmer, J. Petersen, D. B. Sivyer, C. Mills, E. Young, and E. R. Parker,

“Impact of long-term benthic trawl disturbance on sediment sorting and

biogeochemistry in the southern North Sea,” Mar. Ecol. Prog. Ser., vol. 298,

pp. 79–94, 2005.

[40] D. Leo, M. Gauthier, J. Nephin, S. Mihály, and S. K. Juniper, “Bottom trawling

and oxygen minimum zone influences on continental slope benthic community

structure off Vancouver Island (NE Pacific),” Deep. Res. Part II, vol. 137, pp.

404–419, 2017.

[41] F. Olsgard, M. T. Schaanning, S. Widdicombe, M. A. Kendall, and M. C.

Austen, “Effects of bottom trawling on ecosystem functioning,” J. Exp. Mar.

Bio. Ecol., vol. 366, pp. 123–133, 2008.

[42] D. E. Duplisea, S. Jennings, S. J. Malcolm, R. Parker, and D. B. Sivyer,

“Modelling potential impacts of bottom trawl fisheries on soft sediment

51

biogeochemistry in the North Sea,” Geochem. Trans., vol. 14, pp. 1–6, 2001.

[43] C. Dounas et al., “Large-scale impacts of bottom trawling on shelf primary

productivity,” Cont. Shelf Res., vol. 27, pp. 2198–2210, 2007.

[44] C. Bradshaw et al., “Bottom trawling resuspends sediment and releases

bioavailable contaminants in a polluted fjord,” Environ. Pollut., vol. 170, pp.

232–241, 2012.

[45] M. L. Brennan et al., “Quantification of bottom trawl fishing damage to ancient

shipwreck sites,” Mar. Geol., vol. 371, pp. 82–88, 2016.

[46] J. S. Krumholz and M. L. Brennan, “Fishing for common ground:

Investigations of the impact of trawling on ancient shipwreck sites uncovers a

potential for management synergy,” Mar. Policy, vol. 61, pp. 127–133, 2015.

[47] L. Watling and E. A. Norse, “Disturbance of the Seabed by Mobile Fishing

Gear: A Comparison to Forest Clearcutting.,” Conserv. Biol., no. 12, pp. 1180–

1197, 1998.

[48] S. Jennings, M. D. Nicholson, T. A. Dinmore, and J. E. Lancaster, “Effects of

chronic trawling disturbance on the production of infaunal communities,” Mar.

Ecol. Prog. Ser., vol. 243, pp. 251–260, 2002.

[49] European Parliament, “Directive 2008/56/EC of the European Parliament and

of the council of 17 June 2008 establishing a framework for community action

in the field of marine environmental policy (Marine Strategy Framework

Directive),” Off. J. Eur. Union, pp. 19–40, 2008.

[50] T. A. Dinmore, D. E. Duplisea, B. D. Rackham, D. L. Maxwell, and S.

Jennings, “Impact of a large-scale area closure on patterns of fishing

disturbance and the consequences for benthic communities,” ICES J. Mar. Sci.,

vol. 60, pp. 371–380, 2003.

52

[51] N. T. Hintzen, G. J. Piet, and T. Brunel, “Improved estimation of trawling

tracks using cubic Hermite spline interpolation of position registration data,”

Fish. Res., vol. 101, pp. 108–115, 2010.

[52] J. Lee, L. C. Limited, and A. South, “Developing reliable , repeatable and

accessible methods to provide high- resolution estimates of fishing-effort

distributions from vessel monitoring system ( VMS ) data,” ICES J. Mar. Sci.

[53] C. M. Mills, S. E. Townsend, S. Jennings, P. D. Eastwood, and C. A.

Houghton, “Estimating high resolution trawl fishing effort from satellite-based

vessel monitoring system data,” ICES J. Mar. Sci., vol. 64, no. 2, pp. 248–255,

2007.

[54] K. L. Skaar, B. K. H. Ulvestad, and A. Enga, “Short communication Accuracy

of VMS data from Norwegian demersal stern trawlers for estimating trawled

areas in the Barents Sea,” ICES J. Mar. Sci., vol. 68, no. 8, pp. 1615–1620,

2011.

[55] M. J. Witt and B. J. Godley, “A Step Towards Seascape Scale Conservation :

Using Vessel Monitoring Systems ( VMS ) to Map Fishing Activity,” PLoS

One, vol. 2, no. 10, 2007.

[56] L. Buhl-Mortensen, K. Ellingsen, P. Buhl-Mortensen, K. L Skaar, G.

Gonzalez-Mirelis, and K. L Gonzalez-Mirelis, “Trawling disturbance on

megabenthos and sediment in the Barents Sea: Chronic effects on density,

diversity, and composition,” ICES J. Mar. Sci., vol. 73, pp. i98–i114, 2015.

[57] M. Sciberras et al., “Impacts of bottom fishing on the sediment infaunal

community and biogeochemistry of cohesive and non-cohesive sediments,”

Limnol. Oceanogr., vol. 61, no. 6, pp. 2076–2089, 2016.

[58] O. R. Eigaard et al., “Estimating seabed pressure from demersal trawls, seines,

53

and dredges based on gear design and dimensions,” ICES J. Mar. Sci., vol. 73,

pp. i27–i43, 2016.

[59] L. Buhl-Mortensen, K. Ellingsen, P. Buhl-Mortensen, K. L Skaar, G.

Gonzalez-Mirelis, and K. L Gonzalez-Mirelis, “Trawling disturbance on

megabenthos and sediment in the Barents Sea: chronic effects on density,

diversity, and composition,” ICES J. Mar. Sci., vol. 73, pp. i98–i114, 2015.

[60] A. J. Kenny et al., “An overview of seabed mapping technologies in the

context of marine habitat classification.,” ICES J. Mar. Sci., vol. 60, pp. 411–

418, 2003.

[61] A. V. Lazǎr, A. D. Bugheanu, G. V. Ungureanu, and A. D. Ionescu, “Side-scan

sonar mapping of anthropogenically influenced seafloor: A case-study of

Mangalia Harbor, Romania,” Geo-Eco-Marina, vol. 19, pp. 59–64, 2013.

[62] E. Fakiris and G. Papatheodorou, “Calibration of Textural Analysis Parameters

Towards Valid Classification of Sidescan Sonar Imagery .,” In Proceedings of

the 2nd International Conference & Exhibition on “Underwater Acoustic

Measurements: Technologies & Results” Papadakis and Bjorno, Crete, Greece,

25–29 June 2007; pp. 1253–1264.

[63] A. Savini, “Side-Scan Sonar as a Tool for Seafloor Imagery : Examples from

the Mediterranean Continental Margin,” IntechOpen, 2011.

[64] J. M. Bell, “A Model for the Simulation of Sidescan Sonar,” Heriot-Watt

University Department, 1995.

[65] X. Wang, J. Zhao, B. Zhu, T. Jiang, and T. Qin, “A Side Scan Sonar Image

Target Detection Algorithm Based on a Neutrosophic Set and Diffusion Maps,”

Remote Sens., vol. 10, p. 295, 2018.

[66] D. Tamsett and P. Hogarth, “Sidescan Sonar Beam Function and Seabed

54

Backscatter Functions from Trace Amplitude and Vehicle Roll Data,” IEEE J.

Ocean. Eng., vol. 41, no. 1, pp. 155–163, 2016.

[67] I. Klaucke, “Sidescan Sonar,” in Submarine Geomorphology, 2018, pp. 13–25.

[68] M.-J. Patsourakis, E. Fakiris, and G. Papatheodorou, “A method to estimate

otter trawling effects on the seafloor morphology : A case study from Gulf of

Patras , Greece .,” in 13th Pan-Hellenic conference of the Greek

ichthyologists.At: Lesvos - Greece., 27–30 September 2007.

[69] M. Demestre et al., “The need for fine-scale assessment of trawl fishing effort

to inform on an ecosystem approach to fisheries: Exploring three data sources

in Mediterranean trawling grounds,” Mar. Policy, vol. 62, pp. 134–143, 2015.

[70] I. D. Tuck, S. J. Hall, M. R. Robertson, E. Armstrong, and D. J. Basford,

“Effects of physical trawling disturbance in a previously unfished sheltered

Scottish sea loch,” Mar. Ecol. Prog. Ser., vol. 162, pp. 227–242, 1998.

[71] J. DeAlteris, L. Skrobe, and C. Lipsky, “The significance of seabed disturbance

by mobile fishing gear relative to natural processes: a case study in

Narragansett Bay, Rhode Island.,” Am. Fish. Soc., vol. 22, pp. 224–237, 1999.

[72] T. M. Dellapenna, M. A. Allison, G. A. Gill, R. D. Lehman, and K. W.

Warnken, “The impact of shrimp trawling and associated sediment

resuspension in mud dominated, shallow estuaries,” Estuar. Coast. Shelf Sci.,

vol. 69, no. 3–4, pp. 519–530, 2006.

[73] S. De Juan, C. Lo Iacono, and M. Demestre, “Benthic habitat characterisation

of soft-bottom continental shelves: Integration of acoustic surveys, benthic

samples and trawling disturbance intensity,” Estuar. Coast. Shelf Sci., vol. 117,

pp. 199–209, 2013.

[74] S. De Juan, M. Demestre, and P. Sanchez, “Exploring the degree of trawling

55

disturbance by the analysis of benthic communities ranging from a heavily

exploited fishing ground to an undisturbed area in the NW Mediterranean,” Sci.

Mar., vol. 75, no. 3, pp. 507–516, 2011.

[75] C. J. Smith, A. C. Banks, and K. N. Papadopoulou, “Improving the quantitative

estimation of trawling impacts from sidescan-sonar and underwater-video

imagery,” ICES J. Mar. Sci., vol. 64, pp. 1692–1701, 2007.

[76] A. Lucchetti, A. Sala, and J. M. Jech, “Impact and performance of

Mediterranean fishing gear by side-scan sonar technology,” Can. J. Fish.

Aquat. Sci., vol. 69, pp. 1806–1816, 2012.

[77] A. Palanques, P. Puig, J. Guillén, M. Demestre, and J. Martín, “Effects of

bottom trawling on the Ebro continental shelf sedimentary system (NW

Mediterranean),” Cont. Shelf Res., vol. 72, pp. 83–98, 2014.

[78] P. Krost, M. Bernhard, F. Werner, and W. Hukriede, “Otter trawl tracks in Kiel

Bay (Western Baltic) mapped by side-scan sonar,” Meeresforsch, vol. 32, pp.

344–353, 1990.

[79] P. Sanchez, “The impact of otter trawling on mud communities in the

northwestern Mediterranean,” ICES J. Mar. Sci., vol. 57, pp. 1352–1358, 2000.

[80] B. G. González, Y. Petillot, and C. Smith, “Detection and classification of

trawling marks in side scan sonar images,” In Proceedings of the Advances in

Technology for Underwater Vehicles Conference, ExCeL, London, UK, 16–17

March, 2004.

[81] J. Canny, “A Computational Approach to Edge Detection,” IEEE Trans.

Pattern Anal. Mach. Intell., vol. 8, no. 6, pp. 679–698, 1986.

[82] Y. Tang, X. Yu, and N. Wang, “Trawling Pattern Analysis with Neural

Classifier,” in Advances in Natural Computation; Second International

56

Conference, 2006, pp. 598–601.

[83] M. A. Malik and L. A. Mayer, “Investigation of seabed fishing impacts on

benthic structure using multi-beam sonar, sidescan sonar, and video,” in U.S.

Hydrographic Conference, San Diego, CA, USA, 29–31 March, 2005.

[84] T. Sams, J. L. Hansen, E. Thisen, and B. Stage, “Segmentation of Sidescan

Sonar Images,” Danish Defence Research Establishment, Copenhagen,

Denmark, 2004.

[85] P. Blondel, L. M. Parson, and V. Robigou, “TexAn: textural analysis of

sidescan sonar imagery and generic seafloor characterisation,” in IEEE

Oceanic Engineering Society. OCEANS’98. Conference Proceedings (Cat.

No.98CH36259), 1998, vol. 1, pp. 419–423.

[86] E. Fakiris and G. Papatheodorou, “Sonar Class: A MATLAB toolbox for the

classification of side scan sonar imagery, using local textural and

reverberational characteristics,” in International conference on underwater

acoustic measurements (UAM), Nafplion, Greece, 21–26 June 2009, pp. 1445–

1450.

[87] QTC (Quester Tangent Corporation), “QTC Multiview, Acoustic Seabed

Classification for Multibeam Sonar.”, QTC: Sidney, Canada, 2005.

[88] M. Tuceryan and A. K. Jain, “Texture Analysis,” in The Handbook of Pattern

Recognition and Computer Vision, C. H. Chen, L. F. Pau, and P. S. P. Wang,

Eds. River Edge, NJ, USA: World Scientific Publishing Co., Inc., 1993, pp.

235-276. ISBN 981-02-1136-8.

[89] R. M. Haralick, “Statistical and structural approaches to texture,” Proc. IEEE,

vol. 67, no. 5, pp. 786–804, 1979.

[90] E. Fakiris and G. Papatheodorou, “Quantification of regions of interest in swath

57

sonar backscatter images using grey-level and shape geometry descriptors: The

TargAn software,” Mar. Geophys. Res., vol. 33, no. 2, pp. 169–183, 2012.

[91] G. R. Cross and A. K. Jain, “Markov Random Field Texture Models,” IEEE

Trans. Pattern Anal. Mach. Intell., vol. PAMI-5, pp. 25–39, 1983.

[92] B. J. Super and A. C. Bovr, “Localized Measurement of Image Fractal

Dimension Using Gabor Filters,” J. Vis. Commun. Image Represent., vol. 2, no.

2, pp. 114–128, 1991.

[93] O. Daniell, Y. Petillot, and S. Reed, “Unsupervised sea-floor classification for

automatic target recognition,” in International Conference on Underwater

Remote Sensing, 2012.

[94] R. Bajcsy and L. Lieberman, “Texture Gradient as a Depth Cue,” Comput.

Graph. image Process., vol. 5, no. 1, pp. 52–67, 1976.

[95] G. Ferentinos, M. Brooks, and T. Doutsos, “Quaternary tectonics in the Gulf of

Patras, western Greece,” J. Struct. Geol., vol. 7, no. 6, pp. 713–717, 1985.

[96] D. Piper and A. Panagos, “Surficial sediments of the Gulf of Patras,”

Thalassographica, vol. 3, no. 1, pp. 5–20, 1980.

[97] G. Papatheodorou, T. Hasiotis, and G. Ferentinos, “Gas-charged sediments in

the Aegean and Ionian Seas, Greece,” Mar. Geol., vol. 112, pp. 171–184, 1993.

[98] T. Hasiotis, G. Papatheodorou, N. Kastanos, and G. Ferentinos, “A pockmark

field in the Patras Gulf (Greece) and its activation during the 14/7/93 seismic

event,” Mar. Geol., vol. 130, no. 3–4, pp. 333–344, 1996.

[99] A. Lascaratos, “Wind-induced upwellings and currents in the gulfs of Patras,

Nafpaktos and Korinthos, Western Greece,” Oceanol. Acta, vol. 12, no. 3, pp.

159–164, 1988.

[100] I. Maina, S. Kavadas, S. Katsanevakis, S. Somarakis, G. Tserpes, and S.

58

Georgakarakos, “A methodological approach to identify fishing grounds: A

case study on Greek trawlers,” Fish. Res., vol. 183, pp. 326–339, 2016.

[101] P. Throckmorton, H. E. Edgerton, and E. Yalouris, “The Battle of Lepanto

search and survey mission (Greece), 1971–72,” Int. J. Naut. Archaeol., vol. 2,

no. 1, pp. 121–130, 1973.

[102] M. Georgiadis et al., “Coralligène formations in the eastern Mediterranean Sea:

Morphology, distribution, mapping and relation to fisheries in the southern

Aegean Sea (Greece) based on high-resolution acoustics,” J. Exp. Mar. Bio.

Ecol., vol. 368, pp. 44–58, 2009.

[103] E. Fakiris, D. Williams, M. Couillard, and W. Fox, “Sea-floor acoustic

anisotropy and complexity assessment towards prediction of ATR

performance,” Proc. Int. Conf. Exhib. Underw. Acoust., Corfu, Greece, 23–28

June 2013, pp. 1277–1284, 2013.

[104] M. S. Al-Rawi et al., “Intensity Normalization of Sidescan Sonar Imagery,”

2016 Sixth Int. Conf. Image Process. Theory, Tools Appl., 2016.

[105] Y.-C. Chang, S. Hsu, and C. Tsai, “Sidescan sonar image processing:

Correcting brightness variation and patching gaps,” J. Mar. Sci. Technol., vol.

18, no. 6, pp. 785–789, 2010.

[106] A. Burguera and G. Oliver, “High-resolution underwater mapping using Side-

Scan Sonar,” PLoS One, vol. 11, no. 1, 2016.

[107] M. Rahnemoonfar, M. Yari, A. Rahman, and R. Kline, “The First Automatic

Method for Mapping the Pothole in Seagrass,” in IEEE Computer Society

Conference on Computer Vision and Pattern Recognition Workshops, 2017, pp.

267–274.

[108] R. Garcia, T. Nicosevici, and X. Cufi, “On the way to solve lighting problems

59

in underwater imaging,” in Oceans ’02 MTS/IEEE, 2002, vol. 2, pp. 1018–

1024.

[109] J. R. Parker, Algorithms for Image Processing and Computer Vision, 2nd ed.

Wiley Publishing, 2010.

[110] I. Mandhouj, H. Amiri, F. Maussang, and B. Solaiman, “Sonar image

processing for underwater object detection based on high resolution system,” in

Second Workshop on Signal and Document Processing, 2012.

[111] U. Anitha and S. Malarkkan, “A Novel Approach for Despeckling of Sonar

Image,” Indian J. Sci. Technol., vol. 8, pp. 252–259, 2015.

[112] W. Tian, “Side-scan sonar techniques for the characterization of physical

properties of artificial benthic habitats.,” Brazilian J. Oceanogr., vol. 59, pp.

77–90, 2011.

[113] D. Wilken, P. Feldens, T. Wunderlich, and C. Heinrich, “Application of 2D

Fourier filtering for elimination of stripe noise in side-scan sonar mosaics,”

Geo-Marine Lett., vol. 32, no. 4, pp. 337–347, 2012.

[114] C. Tomasi and R. Manduchi, “Bilateral filtering for gray and color images,” in

Sixth International Conference on Computer Vision (IEEE Cat.

No.98CH36271), 1998, pp. 839–846.

[115] F. Durand and J. Dorsey, “Fast bilateral filtering for the display of high-

dynamic-range images,” in Proceedings of the 29th annual conference on

Computer graphics and interactive techniques - SIGGRAPH ’02, 2002, pp.

257–266.

[116] P. Viola and M. Jones, “Robust real-time face detection,” Int. J. Comput. Vis.,

vol. 57, no. 2, pp. 137–154, 2004.

[117] E. Fakiris et al., “Object-based classification of sub-bottom profiling data for

60

benthic habitat mapping. Comparison with sidescan and RoxAnn in a Greek

shallow-water habitat,” Estuar. Coast. Shelf Sci., vol. 208, pp. 219–234, 2018.

[118] E. Fakiris et al., “Multi-Frequency , Multi-Sonar Mapping of Shallow Habitats

— Efficacy and Management Implications in the National Marine Park of

Zakynthos , Greece,” Remote Sens., vol. 11, no. 4, 2019.

[119] D. P. Williams and E. Fakiris, “Exploiting Environmental Information for

Improved Underwater Target Classification in Sonar Imagery,” IEEE Trans.

Geosci. Remote Sens., vol. 52, no. 10, pp. 6284–6297, 2014.

[120] D. Bradley and G. Roth, “Adaptive Thresholding Using the Integral Image,” J.

Graph. Tools, vol. 12, no. 2, pp. 13–21, 2007.

[121] N. Modalavalasa, G. S. Rao, and K. S. Prasad, “A Novel Approach for

Segmentation of Sector Scan Sonar Images using Adaptive Thresholding.,” Int.

J. Inf. Sci. Tech., vol. 2, no. 4, pp. 113–119, 2012.

[122] S. Banerjee, R. Ray, S. N. Shome, and G. Sanyal, “Noise induced feature

enhancement and object segmentation of forward looking SONAR Image,”

Procedia Technol., vol. 14, pp. 125–132, 2014.

[123] J. R. Parker, “Gray level thresholding in badly illuminated images,” IEEE

Trans. Pattern Anal. Mach. Intell., vol. 13, no. 8, pp. 813–819, 1991.

[124] R. A. Coggan et al., “Comparison of rapid methodologies for quantifying

environmental impacts of otter trawls,” 2001.

[125] D. Tamsett, “Geometrical Spreading Correction in Sidescan Sonar Seabed

Imaging,” J. Mar. Sci. Eng., vol. 5, no. 4, p. 54, 2017.

[126] T. B. S. Correa, M. Grasmueck, G. P. Eberli, K. Verwer, and S. J. Purkis,

“Deep Acoustic Applications,” in Coral Reef Remote Sensing, J. A. Goodman,

S. J. Purkis, and S. R. Phinn, Eds. London, UK: Springer, 2013, pp. 253–282.

61

[127] C. Peignon, M. Kayal, E. Poisson, M. Adjeroud, and L. Penin, “Spatial Patterns

and Short-term Changes of Coral Assemblages Along a Cross-shelf Gradient in

the Southwestern Lagoon of New Caledonia,” Diversity, vol. 11, no. 2, p. 21,

2019.

[128] M. A. Morrison, E. G. Jones, M. Consalvey, and K. Berkenbusch, “Linking

marine fisheries species to biogenic habitats in New Zealand : a review and

synthesis of knowledge,” 2014.

62

Appendix

63

Below are the article entitled: "Automatic detection of trawl-marks in sidescan sonar images through spatial domain filtering, employing Haar-like features and morphological operations" that has been published in the MDPI Geosciences journal and the successfully submitted abstract of the paper entitled “Environmentally adaptive automatic detection of linear seafloor features in sidescan sonar imagery: The case of trawl marks detection” for the 5th international conference on Underwater Acoustics.

Gournia, C.; Fakiris, E.; Geraga, M.; Williams, D.P.; Papatheodorou, G.

Automatic Detection of Trawl-Marks in Sidescan Sonar Images through Spatial

Domain Filtering, Employing Haar-Like Features and Morphological Operations.

Geosciences 2019, 9, 214

https://www.mdpi.com/2076-3263/9/5/214

64

65

This paper makes a detailed introduction to the design of a new automatic linear seafloor feature detection algorithm and its implementation on sidescan sonar (SSS) records, with its focus on trawl marks (TMs) detection. TMs are long linear scars on the seafloor, which are products of bottom trawl fishing. In image processing, detection of lines is a classical problem that seems to be more challenging on underwater acoustic imaging as the line segments are intermixed with wide-ranging environment backgrounds, and acoustic radiometric and geometric artifacts. Therefore, classic edge detection techniques based on the intensity gradient of the image do not yield reliable detection on SSS images. This proposed method integrates the characteristics of the linear features of interest in an environmentally adaptive procedure and is divided into three major steps. At the first step, preprocessing image techniques are applied to the original images. In the main stage, a spatial-domain filter is implemented through multi- scale rotated Haar-like features and integral images that measures the level of multiple oriented contrasts between adjacent areas. Seafloor characterization based on

Anisotropy and Complexity calculations over the Haar-like filter’s responses identifies three types of seafloor texture: complex (e.g. biogenic mounds, clutter), anisotropic

(e.g. TMs, ripples), and plain (e.g. undisturbed sand). At the same step, another function over filter’s responses produces a map that highlights the accurate locations where the candidate linear features prevail. The produced map is automatically binarized, morphologically processed and every linear image object is undergone properties measurement. The final linear features are selected according to a set of geometric and background textural feature criteria. In this study, is presented a set of assignment criteria that is tailored to the specific needs of TMs detection and is followed by TMs quantification that provides valuable measures for the estimation of bottom trawling impacts.

66