ε. . Εστ. Έδ. ΕΕ, Ι, τε. 2-3 2008 Tech. Chron. Sci. J. TCG, I, No 1-2 37 Summary Photointerpretation of linear information is a subjective process and therefore there is a substantial need for automation of extracting linear information using automated techniques. Certain efforts were made in this direction including the application of edge enhancement and detection operators, wavelets, Hough Transform etc. However it is diffcult to choose among optimal algorithms since the complex scenes portrayed on satellite images are strongly dependent on the radiometric and physical properties of the sensors and on the illumination properties and topographic relief of each scene. Therefore, the category of information to be extracted (scale and context) determines the “suitability” of the method applied for linear feature extraction. In this context, the objective of this work was the implementation, evaluation and comparison of selected optimal edge detection algorithms combined with complementary remote sensing methods towards automated linear feature extraction (roads, land use boundaries, buildings, etc) in an urban / peri-urban environment. The test areas used were located in the extended area Attica Prefecture. A multispectral IKONOS image and panchromatic KVR-1000 imagery of two different dates were acquired for the case studies of Aghios Stefanos and Penteli, Attica, Greece respectively and processed by the following edge detectors: (a) The Canny edge detection algorithm, (b) The Rothwell algorithm, (c) The LOG-LIN algorithm, (d) The SUSAN operator, (e) The anisotropic diffusion algorithm of Black, (f) The Bezdek algorithm and (g) The EDISON algorithm. The resulted edge maps were then compared to thematic maps resulted by applied remote sensing methods and techniques and assessed using statistical methodology. Finally, the performance and behavior of each algorithm for urban feature extraction was assessed as well as the suffciency of the edge detection methodol- ogy for environmental change detection purposes. 1. INTRODUCTION The urban and peri-urban environment is a widely inves- tigated research topic among geo-scientists, as well as a very crucial subject for enforcing administrative and legal poli- cies. Nowadays, new geo-data resources from high spatial resolution remote sensing sensors (e.g. IKONOS, Quickbird, KVR, etc) have the potential to improve mapping and analy- sis of urban land cover / land use structures and to monitor related dynamics. Remote Sensing methodology can provide very powerful tools and means for monitoring the change dynamics of the urban environment by exploiting the spec- tral and spatial information and context of these geo-data and for further decision-making. For many years, edge detection has been proved a very valuable tool for automating the extraction of information at a low level. The aim of edge detection is to extract features, as meaningful as possible depending on the physical illumi- nation properties of an image. Edge detection techniques have been successfully applied in the domains of Computer Vision (Haralick, 1984, Canny, 1986, etc), Biomedicine (Bezdek, et.al., 1998), Geomorphology (Argialas and Ma- vrantza, 2004), etc. In this work, the application and performance of sophis- ticated and automated edge detection techniques is assessed using spectral and spatial information from high resolution images in an urban / peri-urban environment, as a stand- alone methodology and as a combined methodology with other remote sensing methods and techniques for extracting urban features of interest (building boundaries, road seg- ments, etc). 1.1 Edge detection algorithms: An overview In image processing and computer vision, edge detec- tion treats the localization of signifcant variations of a gray level image and the identifcation of the physical and geo- metrical properties of objects of the scene. The variations in the gray level image, commonly include discontinuities (step edges), local extrema (line edges) and junctions. Most The Role of Edge Detection Techniques for the Extraction of Linear Information in Urban / Peri-Urban Environment OURANIA D. MAVRANTZA DEMETRE P. ARGIALAS Professor, National Technical University of Athens