Image Enhancement for Vegetative Pattern Change Analysis

CENE L. BROTHERS ERNEST B. FISH Texas Tech University Lubbock, TX 79409 Image Enhancement for Vegetative Pattern Change Analysis Photographic enhancement-overlay processing of aerial imagery gave improved results when compared with traditional change detection techniques. INTRODUCTION location of change, (3) the magnitude of ATURAL RESOURCES have always been of change, (4) the direction of change, and (5) N vital importance to man, and recent in variations in the rates of change (Estes et al., creasing demands on them have progres 1972). Time spans between sequential data sively intensified the need for improved collection dictate the degree of accuracy to management (Carneggie and DeCloria, 1973; which a manager can determine these as Shepard, 1964). Monitoring changes in ele pects ofchange. An effective monitoring sys ments of concern to resource managers con tem would be one providing relevant, non situtes an important phase in the manage- redundant information in an efficient and ABSTRACT: Valid management decision making in natural resource planning is dependent upon an accurate data base and the ability to detect changes in resource status. Photographic enhancement overlay processing of aerial imagery for change detection was de veloped and tested. Compared with traditional change detection techniques, the new method gave improved results in terms ofgen eral accuracy and the specific nature ofchanges occurring where the objects of interest could be isolated to a specific optical density range. Efficiency of the technique is directly affected by imagery scale and size ofarea to be monitored. ment process. Changes reflect how the envi cost effective manner which can be readily ronment is reacting to management practices used with a sustained management infor and, thus, provide input for an evaluation of mation system (Paul, 1976; Shepard, 1964). the practices. Change detection of resources has been Most sequential monitoring techniques carried out in various subfields of resource contain the major disadvantage that, unless management in attempts to obtain broad the data are collected while the change is ideas of resource conditions. Studies have actually occurring, the exact time of the shown aerial photography to be superior to change cannot be documented but only other data collection methods because it al speculated about within the time span ofthe lows assessment of a complete area and pro sequential collection of data (EI-Ashry and vides an accurate recording of surface situa Wanless, 1967). However, sequential tions (Collins, 1972; Driscoll, n.d.; Shaklee, monitoring can provide managers with in 1976). Studies reviewed here were based on formation on (1) the nature ofchange, (2) the aerial photographic data collection; how- PHOTOGRAMMETRIC ENGINEERING AND REMOTE SENSING, 607 Vol. 44, No.5, May 1978, pp. 607-616. 608 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING, 1978 ever, various analysis and change detection tronic density scanner. Values computed re procedures were employed. suit in an "index" which increases as differ ences in vegetation on sequential photog LITERATURE REVIEW raphy increase (Rouse et al., 1973). Traditional interpretation and mapping Frequently, resource managers are over techniques seem to be most frequently used whelmed by the expense and sophistication in studies involving change detection. of some of these change detection tech Baseline conditions are interpreted and niques and conclude that they cannot be mapped from imagery which is then com of assistance to a particular management pared with information drawn from sequen problem. However, demand for rapid collec tial imagery of the same area. Measures of tion of information exists, and aerial photog the total area covered by a resource of con raphy must be considered as a means to cern are taken from both sets of aerial pho provide administrators with the required tography. This information is then compared understanding or with the capacity to effi to determine change in total area (Avery, ciently acquire an understanding of re 1965; Bubenik, 1976; Collins, 1972; Estes et sources (Carneggie, 1970; Poulton, 1970). al., 1972; Richter, 1969; Wray, 1971). Loca One of the main limitations of interpreta tions of change have been determined from tion is the restricted ability ofthe human eye sequential data by overlaying maps drawn to readily recognize subtle differences in the from the imagery at the same scale or by pro diagnostic characteristics of images jection of an image onto a baseline map (Nielsen, 1974). Image enhancement (Milazzo and Lins, 1972; Olson and Meyer, techniques have been developed which re 1976; Shaklee, 1976; Wagner, 1963). These duce this restriction through alteration ofthe traditional interpretive techniques are image, accentuating slight differences which time-consuming, wearisome, and subject to could be of interest to the observer. There omission enors (Shepard, 1964). are primarily two enhancement processes; Techniques have been developed which electronic, which converts image densities consider the problems of traditional in to voltage, and photographic, which alters an terpretation and attempt to reduce amounts image by re-exposing it on photographic of interpretation needed to determine film. change. Sequential images can be subtracted Electronic processes which can change or averaged with one another by overlaying image densities to voltage display densities, positive and negative transparencies of an on a cathode-ray tube, allow for various en area in register. This results in an image in hancements of an image. Density-slicing or which areas that have not changed appear the separation of density ranges is the basis medium gray while areas of change appear upon which this process works. Density as light or dark tones (Jones et al., 1975; ranges are imaged as separate colors, dis Reeves, 1975; Simonett, 1974). A "blink" played all at once or separately on a televi process permits viewing of sequential im sion screen (Sayn-Wittgenstein, 1973; ages alternately at about ten frames per sec Schlosser, 1974). ond. Areas on the image which "blink" as Several photographic processes which the pictures are alternated represent areas have been employed to assist the interpreter which have changed (Masry et al., 1975; in identifying subtle image differences in Simonett, 1974). These techniques are lim clude: tone contrast stretching, posteriza ited, in that changes of interest to policy tion, edge isolation, Sabattier effect, Ag makers may be embedded in a "larger facontour density-slicing, and litho density amount of extraneous change, or clutter" slicing (Clarke, 1967; Kodak, 1973; Nielsen, (Simonett, 1974: 64). 1972, 1974; Rodriguez-Berjarano, 1975; Electronic scanning equipment also has Ross, 1976; Simonett, 1974). been employed to reduce traditional in Image enhancement provides a technique terpretation needed to detect change. A which can be used to isolate images of peripheral processor scanner was utilized as selected resources. Isolation ofthese images a "quick-look" change detection technique. can be a means ofelimination ofthe problem The scanner records edges of objects on se of "clutter" effecting interpretation of quential photographs that are subsequently changes discussed by Simonett (1974). Thus, compared by using a computer (Humiston a combination of an enhancement process to and Tisdale, 1973). Density changes of isolate images and an overlay process of natural vegetation on sequential imagery positive and negative transparencies of se have been detected by employing an elec- quential imagery to detect change should IMAGE ENHANCEMENT FOR VEGETATIVE PATTERN CHANGE 609 provide resource managers with a method this process was tested using model situa which gives information rapidly and can be tions and sequential aerial photography. utilized easily. A density step chart was employed to Many changes ofinterest to land managers model objects of various tones.* Exposures have a direct effect on vegetation and result of the density chaIt were made onto ortho in alteration of vegetative patterns. Plant type film with exposures beginning at 2.5 communities integrate separate growth fac seconds and at 2.5 second intervals up to a tors which affect the community, and pro final exposure of 10 seconds. These nega vide a biological expression of these fac tives were then contact printed on ortho film tors through the structure and components of to produce positives. Resultant positives and the community (Poulton, 1970). Vegetation negatives of different exposures were over has been used in numerous studies to laid in order to mask specific density ranges. monitor ana detect significant resource con Simulation of aerial photography was ditions including urba.n, wildlife, geologic, achieved through use of near vertical photo and ecologic analyses (Collins, 1972; Dill, graphs taken of a three-dimensional land 1963; Driscoll, n.d.; Ducker and Horton, scape model. An initial picture was taken of 1971; Estes et ai., 1972; Hanson and Smith, the model to represent a baseline condition. 1970; Jones et ai., 1975; Larson, 1967; A second photograph was taken after several Lewis, 1974; Morrison, 1972). changes had been made in the simulated Presently, the gap which exists between trees. Exposures ofthe initial negatives were capabilities for gathering extensive amounts made on ortho film beginning at a 5 second ofaerial imagery and the limited capabilities exposure and

Load more