期刊名称:ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN:2194-9042
电子版ISSN:2194-9050
出版年度:2006
卷号:XXXVI Part 4
出版社:Copernicus Publications
摘要:In the satellite images, built up area is manifested as texture. Therefore, built up area analysis can be approached using texture concepts. It is observed in many gray scale high resolution satellite images that building tops , ground, vegetation and roads have almost same gray value variations and also sometimes texture of the image surface is not very well defined. So when built up area analysis is performed based on only either gray level variation or texture of the image alone, it does not always give satisfactory results, The proposed approach in this paper involves a multi level processing which uses both texture information as well as tone variation of the image to perform analysis for built up area mapping. The approach is based on Gabor filters and neural networks to extract built up areas in satellite images such as IRS 1C/1D and IKONOS. The main issues addressed in this process are . functional characterization of the filter bank and number of filters, extraction of appropriate texture features from the filtered images, the relationship between filters (dependent vs. independent) and the texture feature extraction. The paper addresses all these issues and proposes a texture feature to reduce the computational time required for Gabor Filter based texture classification. The classification of texture features is done using artificial neural networks. In the second layer of processing, built up area extracted in first layer is segmented using Fuzzy C means clustering resulting in extraction of individual buildings. The output of the approach results in layout of the urban areas which can be used for updating the GIS information or the maps
关键词:Built up ;Texture; Multi Channel Filtering; Gabor Filters; Classification; Neural Net; Morphological P rocessing