MAPPING GEO-TAGGED PICTURES IN A DISASTER MANAGEMENT EVENT
J Masó a, *, P. Diaz a, X. Pons b
a Center for Ecological Reaserch and Forestry Applications (CREAF). Universitat Autònoma de Barcelona, 08193 Bellaterra Barcelona, Spain - (Joan.Maso, Paula.Diaz)@uab.cat b Departament of Geography. Universitat Autònoma de Barcelona, 08193 Bellaterra Barcelona, Spain – [email protected]
Gi4DM2011
KEY WORDS: standards, pictures, geo-tag, map service
ABSTRACT:
The overall objective of GEO-PICTURES (GMES and Earth Observation combined with Position based Image and sensor Communications Technology for Universal Rescue, Emergency and Surveillance) (an FP7 project) is to provide an emergency management solution for global use. Satellite images, in-situ sensors, geo-tagged pictures, texts, etc are integrated on the system, processed in a centralized decision place and, the results, disseminated. Pictures captured on the ground, with position and camera orientation information, are valuable information. This communication explores some approaches and determines the best way to be redistributed them. Geo-tagged pictures can be exposed in four different approaches: Picture URLs, GeoRSS feed, Web Map Service (WMS) or Web Feature Service (WFS). These ways are evaluated in order to provide access to data in an easy, user friendly and robust way. RSS feed can contain pictures and text. GeoRSS adds geo-reference capabilities to RSS and can be easily read by social network tools or Google Earth, but are difficult to integrate with map clients based on OGC-OWS standards like WMS and WFS. WMS Geo- tagged images server defines collection of layers that are portrayed and visualized. All the pictures related to the same event are accumulated in the same layer. WFS Geo-tagged images server serves vector datasets. A simple feature GML profile is used to define the geo-tagged features as a GML points. Mechanisms exposed are explained, evaluated and compared as emergency disaster response tools.
1. INTRODUCTION GEO-PICTURES (GMES and Earth Observation combined with Position based Image and sensor Communications 1.1 General Introduction Technology for Universal Rescue, Emergency and Surveillance) is a European Commission 7th Framework Program project. Unfortunately, catastrophic situation like past earthquake in The overall objective of the projects is to provide a powerful Haiti in January 2010 happens too often. Haiti earthquake was humanitarian and environmental emergency management one of the most destructive events of this century causing solution for global use, represented by concrete user needs from 112,405 deaths and 482,349 homeless people (OCHA Haiti the United Nations, the EU Community Mechanism for Civil situation update map 31 January 2010). The devastation shown Protection and Amazonas, representing the largest by the media around the world generated shock and environmental areas of concern in the world. Several data consternation in the general public. The reaction teams and sources are integrated on the system like satellite images, in-situ humanitarian aid organizations urgently needed information in sensors, geo-tagged pictures, texts, etc. This information are order to coordinate their efforts on the ground. The official processed in a centralized decision place and, the results, response came from several international bodies, such as the disseminated. United Nations Institute for Training and Research (UNITAR) Operational Satellite Applications Programme (UNOSAT) and the International Charter Space and Major Disasters and UN- SPIDER. In addition, unofficial amateur groups with common interests using tools such as OpenStreetMap and Google Mapmaker elaborated cartographic products much faster than the official sources could. These efforts increased the data flow generated in one of the most compelling manifestations of the crowdsourcing phenomena ever seen. Meanwhile, websites collected most of the geospatial information available. These were mainly satellite images and old US thematic maps that were used as base maps to produce new damage assessment maps that could be used to respond immediately to the crisis.
* Corresponding author.
tagged field photos are used to document disease outbreak locations, bird survey, plant survey, and soil samples. One example is the Global Geo-Referenced Field Photo Library (Xiao, 2010) where users to upload geo-tagged photos, query, interpret, classify, and download photos and resultant thematic database of land cover types, and then use the database for GIS analysis. Another example is the Geo-Wiki project where geo- tagged pictures are used to validate global land cover maps (Fritz 2008). Angermann (2009) and Frassl (2010) apply geo- tagged pictures also to disaster management obtained from a camera that is linked to a specific communication suitcase in a decentralized disaster management schema that shares some point in common with the GEO-PICTURES general solution.
GEO-PICTURES provides a tools that easily send the images to a centralized decision server where experts analyze them and assessing the situation and releasing some digested information
Figure 1: GEO-PICTURES general objectives back to the field. The ways that a subset of the geo-tagged images has to be redistributed is what this paper discusses and The main objective of the GEO-PICTURES project is to determines the best of them. provide end users with all the required information in a timely, seamlessly integrated, secure and user-friendly way for 1.2.1 Geo-tagged RSS feeds managing disasters and emergencies. Solutions developed in the RSS feed is a popular technology to publicize news on project will help to rapidly collect in-situ information from particular events. A RSS feed is an XML encoded file emergency field assessments. This information are integrated composed by a caption and a collection of items. The caption with satellite imagery and make best use of the combined has a minimum set of metadata about the RSS author and the information in a seamless service platform. This information is topic of the items and the set of items are also composed by a then efficiently distributed to the in-field personnel and to the minimum set of metadata (title, author, date of publication, general public (see figure 1). The project will ensure that source, etc) and a link that can get you more information. research and developed will be available as a real solution. Eventually it can contain a reference to a picture. This document is placed in the producer's website in a permanent 1.2 Presenting the data to users URL but the document itself is updated frequently. This way, the user can periodically check the same URL for an updated The data presented to the users consists in updated base maps version with new information. There are several tools that form composed by current satellite images and other cartographic a fixed list of RSS URLs digest this information and show it to products enriched with in-situ data. the user as an integrated and personalized newspaper. The process of adding a new URL to the list of periodically checked Satellite images available in a crisis situation can be distributed URLs is called syndication. (Wick and Becker 2007). The main in several ways. In GEO-PICTURES this images are application of the RSS feeds is to announce the existence of compressed in JPEG2000 format that integrates the wavelet new content and provide small pieces of information that allows compression algorithm that provides high compression rations users to filter them. This property makes it the ideal tool for with minimum quality degradation (Taubman et al. 2002). This distributing new information in a post-event situation. imagery can be directly distributed to the users that have the right decompression tool and can also distributed using WMS GeoRSS adds geo-reference capabilities to RSS. GeoRSS (de la Beaujardiere, 2004) and WMTS (Masó 2010) services, specification defines 3 levels of complexity, W3C-geo, both Open Geospatial Consortium (OGC) international GeoRSS-Simple and GML-GeoRSS. W3C-geo only adds the standards. Each approach has it advantages and disadvantages
GeoTagged images can be automatically generated by a GPS GeoRSS feeds can be easily read by social networks tools such enabled camera, a GPS enabled cell phone or using a software as the Virtual Disaster Viewer (Maiyo 2010), but are difficult to that synchronizes a digital camera with a simultaneous GPS integrate with more traditional map clients based on OGC-OWS receiver record (GPX or NMEA) like GPicSync (Xiao 2010). standards like WMS (Web Map Service) or WFS (Web Feature Service).
The idea of collecting geo-tagged pictures is not new in the literature. Geo-tagged images can be used for several applications and scientists and citizens are using GPS-enabled digital cameras to capture landscapes across the globe. Geo-
2. METHODOLOGY the image in a medium size, a link to the full resolution image and the tags related to the image in an HTML table. 2.1 Different techniques to present geo-tagged pictures 2.1.4 WFS GeoTagged pictures Geo-tagged pictures can be mainly exposed in four different The WFS Geo-tagged images server allows you to serve vector ways: Picture URLs, GeoRSS feed, WMS and WFS (figure 2). datasets. The common (but not the only) use for WFS is to serve These four ways to expose data are evaluated in order to features in GML format. The GetCapabilities operation can use determine the best way to provide access to data in an easy, user a simple feature GML profile to define the geo-tagged features friendly and robust way. as a GML Point. The WFS requires describing the WFS FeatureTypes as a GML application schema do to that, while using the GetFeature operation the user can get a filtered subset Individual Geo-Tagged picture URL of the feature collection of geo-tagged images.
GeoRSS feed 2.2 Geo-Tagged pictures WMS In GEO-PICTURES we identified the need for a WMS server WMS that is able to directly represent geo-tagged images extracted GeoPictures form a RSS feed in a map, in the same way that rapid mapping Server techniques are applied in post-event responses (figure 3).
WFS
Figure 2. Geotagged pictures distribution alternatives.
2.1.1 Individual GeoTagged picture URL The simplest way is directly provide to the user a URL of the geo-tagged image. This approach requires an active attitude from the user that has to visit a web portal from time to time to see if there are new images. URL's do not provide additional metadata so the user will find it difficult to filter by topic or geographic area
2.1.2 GeoRSS with GeoTagged pictures This approach adds the needed metadata to the images without having to download them. When the bandwidth is limited, this characteristic is particularly relevant. Also a RSS aggregator software is able to show the user that new contend has appeared without needing an active user attitude. In GEO-PICTURES, GeoRSS will be used to distribute new images to the in-field Figure 3: Rapid map generated by GAUL, WFP, CNIGS, actors and to the general public. MINUSTAH that shows Port au Prince access roads condition. It is mainly combination of geo-tagged Unfortunately currently, only a few set of web clients are able to pictures map over a base map. present GeoRSS feeds directly to the user and many of them present the pictures with some small icon over a base map This software has the following user requirements without any clue of what is behind the icon. Users can eventually select one of them and get more information and, Picture has to be represented close to the right place particularly, see the picture. Xiao X (2010) shows an example and the exact place must be indicated. of this behaviour. Overlay of the resulting layer with other base maps has to be possible. 2.1.3 WMS GeoTagged pictures A simplification of the picture has to be visible The WMS Geo-tagged images server is an OGC and ISO directly in the resulting image standard defining a collection of layers that can be portrayed Collisions and multiple images render in the same and visualized, as well as optionally queried by location. place have to be avoided. Through the GetCapabilities operation all the pictures related to New pictures must be incorporated as soon as the same event can be accumulated to the same emergency possible. event (identified by an event ID) in the same layer. Through the GetMap operation a layer can be represented in a particular This list of requirements can be reformulated to a more concrete style, two of which are to represent pictures as symbolized point description of what the application has to do. In fact, the WMS maps or as small thumbnails. In order to have more information software has to represent each GeoRSS feed file as a WMS about the image, the GetFeatureInfo returns an HTML page with layer, but this layer has not to be a static one because the application has to check for new versions of the GeoRSS feed
from time to time. The picture data has to be extracted from a case we have seen that there is no speed advantage in pre- picture URL in the GeoRSS feed and the picture position has to rendering the whole space to several resolutions and the result be extracted from the "geo" part of the RSS. Geo-referenced of interpolating some resolutions gives a pour quality result in pictures have to be rendered as icons and as thumbnails. some cases. Renderization of the icons and pictures is done on the flight and the size of the icon (in pixels) remains constant at any scale. In the MiraMon map server extension for GeoRSS we register Where a thumbnail of the picture is presented, the maximum the RSS in the server as a layer in the MiraMon layer collection. size in pixels of these pictures has also to be scale independent. Once the registration is done, an automatic layer process The exact place of the rendered picture can be altered to avoid preparation starts downloading the RSS and converting it to a collisions with other pictures. In this case, a line or an arrow has point layer. The conversion generates several separated point to illustrate the exact position of the picture. Eventually, if the files distributed in vector tiles that will not exceed 512 points number of collisions is high, the software can decide to decrease each. For each point, the original picture is downloaded and the size of the thumbnails to improve visibility. The areas where downgraded to a total width of 128, 64 and 32 pixel thumbnails there is no symbol or thumbnail remains transparent. This and stored in separated files. In the corresponding record of the allows the overlay of this layer with others. database table it saves the path of these thumbnails and also the rest of the RSS item elements as text fields. Several generalized Simple GeoRSS feeds can be seen as point datasets where the point layers are prepared for different resolutions, but this time position of the objects is extracted from the
The MiraMon Map Server WMS implementation was able to serve only raster or vector datasets that has been pre-processed in a way that they have been prerendered in a set of user defined resolutions that are stored in small rectangular raster files that are called tiles. This pre-processing process can be full automated and the only problem is that it can take a long time for a very long datasets. When a WMS client requests one of these resolutions, MiraMon WMS serves data simply by selecting the right tiles, merging them and cutting the result to the right bounding box. If the client requests a resolution that is not prerendered on the server, it is generated by interpolating it form the closest available. Prerendered images have 2 main advantages: speed and quality. In fact, it can serve WMS requests faster than other products because prerendered images are stored in a format closer to the final one and in a fragmentation that allows randomly access to extract the requested subset. Another advantage of this technique is that the low resolution prerendered data can be generated in a way that increases the visual quality of the result by means of introducing good interpolators even if this can take more time to prepare.
Prerender images at different resolutions is a good idea for raster and line and polygon vector formats that has simple Figure 4: Land use map and geo-tagged images shown in a symbology but when we try to represent a point datasets, the WMS client software for Windows Mobile and situations is different. Point datasets use icons and texts to Windows Phone 6.5. It shows the need for a represent the objects in the map and generally, the size of these avoiding collisions in the representation. objects in the screen is the same for each resolution, the image is not fully cover by objects, leaving many void spaces. In this
From time to time (period also decided in the registration ACKNOWLEDGEMENTS process), the server revisits the RSS feed and checks for updates. If an update is detected, the layer preparation process is Authors of this communication thank the support of the executed again European Commission through the FP7-242390-GEO- PICTURES (SPACE-2009-1). Xavier Pons is recipient of an The server has been successfully tested with a collection of ICREA Academia Excellence in Research grant (2011-2015). 7334 images downloaded from www.virtualdisasterviewer.com, most of them coming from the Earthquake Engineering REFERENCES Research Institute. The KML file was transformed to GeoRSS as a test file using XSLT. The layer was successfully included Angermann, M.; Khider, M.; Frassl, M.; Lichtenstern, M.: DMT in the MiraMon server Catalogue and the automatic data (2009) An integrated disaster management tool, First preparation was generated and stored in several point files. The International Conference on Disaster Management and Human WMS layer is able in 2 different styles: thumbnails and static Health Risk: Reducing Risk, Improving Outcomes, New Forest, symbols. Once the server was ready, it was tested using open UK, September 2009 source (such as Openlayers) and proprietary WMS clients de la Beaujardiere, J., 2004 , OGC Web Map Service (WMS) (Intergraph and ESRI) at several resolutions with visual results Interface, Ver.1.3.0, OGC 03-109r1. conformant to the requirements detailed in the previous section. Frassl M., Lichtenstern M., Khider M. and Angermann M. 3.1 Future work (2010) Developing a System for Information Management in Disaster Relief - Methodology and Requirements. Proceedings We plan to support also GeoRSS feed with more elaborated of the 7th International ISCRAM Conference – Seattle, USA, spatial description like lines, tracks (Tai 2009) and polygons. May 2010 Also we will allow the possibility of having more than one picture in the same point. Also new alternatives to improve the Fritz S., McCallum I., Schill C., Perger C., Grillmayer R., implementation of the collision avoidance algorithm will be Achard F., Kraxner F. and Obersteiner M. (2009) Geo- explored. Wiki.Org; The Use of Crowdsourcing to Improve Global Land Cover Remote Sens. 2009, 1, pp 345-354 We are considering the inclusion of camera orientation representation. Orientation information can be introduced by Maiyo L., Kerle N. and Köbben B (2010) Collaborative Post- the user like in www.confluence.org or included in the metadata disaster Damage Mapping via Geo Web Services. Geographic by the internal camera compass. Information and Cartography for Risk and Crisis Management We have experience a problem with the position of the objects Lecture Notes in Geoinformation and Cartography, 2, pp 221- represented in the image, particularly in pictures of the 231, landscape. Sometimes, the pictured object in the photograph is a far away object but the position associated with this object is Masó J, Zabala A and Pons X (2010) Combining JPEG2000 the GPS position. We are considering alternatives based on the Compressed Formats and OGC Standards for Fast and Easy combination on the camera orientation with focus setting or Dissemination of Large Satellite Data. Italian Journal of user intervention to solve this issue. Remote Sensing, 42 (3): pp 101-114
4. CONCLUSIONS Maso J., Pomakis K., Julia N. (2010) OGC Web Map Tile Service (WMTS) Implementation Standard, Ver 1.0.0, OGC 07- GEO-PICTURES results will provide an emergency 057r7. management solution for global use that will use GMES and other satellite images, in-situ sensors, geo-tagged pictures, texts, Pons, X. (2002) MiraMon. Sistema d'Informació Geogràfica i etc, processed in a centralized decision place, filtered and software de Teledetecció. Centre de Recerca Ecològica i redistributed to the in-field aid organization as a rapid mapping Aplicacions Forestals, CREAF. Bellaterra. ISBN: 84-931323-4- products or re-filtered to be disseminated to the mass media and 9 to the public. Pictures captured on the ground, with position and camera Reed C (2006) An Introduction to GeoRSS: A Standards Based orientation information, are valuable information that is Approach for Geo-enabling RSS feeds white paper, Ver 1.0.0. distributed in the form of GeoRSS feeds with URLs to the OGC 06-050r3. actual pictures. The generated GeoRSS can be directly consumed by some clients and social network tools or by Tai C.H., Yang D.N., Lin L.T., and Chen M.S. (2008) Google Maps, but are difficult to integrate with map clients Recommending Personalized Scenic Itinerarywith Geo-Tagged based on OGC-OWS standards like WMS and WFS. A WMS Photos. Multimedia and Expo, 2008 IEEE International extension for GeoRSS with geo-tagged images was developed Conference June 23, pp 1209 - 1212 to allow that. The server design has been prepared to support a Taubman D.S. y Marcellin M.W. (2002) JPEG2000: Image large amount of geo-tagged pictures and requires a preparation compression fundamentals, standards and practice. Kluwer, of a point layer for each resolution. Depending on the number Academic Publishers. of collisions at a particular resolution the layer is generalized into a more simple one that can be fast to deliver as a WMS response and easy to understand. Torniai C. y Battle S. y Cayzer S. (2007) Sharing, Discovering Future work will involve new adding track and polygonal and Browsing Geo-tagged Pictures on the World Wide Web. objects, improve the collision avoidance algorisms and consider The Geospatial Web. Springer pp 105-115 the representation of the camera orientation and object distance.
Wick M. y Becker T (2007) Enhancing RSS Feeds with – Extracted Geospatial Information for Further Processing and Visualization. The Geospatial Web. Springer pp 105-115
Xiao X (2010) Community Remote Sensing: Global Geo- Referenced Field Photo Library, University of Oklahoma. IGARSS2010 Hawai.