Automatic Recognition of Village in Remote Sensing Images by Support Vector Machine Using Co-Occurrence Matrices

Accurate and comprehensive extraction of village is meaningful for a land consolidation project. In consolidation work, it is necessary to recognize village objects quickly and accurately. The high resolution remote sensing images make it possible. However, a highly automatic classification is not easy to attain, because remote sensing images usually contain many complex factors and mixed pixels. The goal of this paper is to develop an automatic classifier for village recognition. Our approach uses a support vector machine (SVM) as a classifier, which tries to detect pixels belonging to villages in the image and reject the others. Texture features are used as inputs to the SVM classifier for identifying villages. The texture analysis is based on the grey-level co-occurrence matrix method. Here several co-occurrence parameters are computed and then compared. Three distinct co-occurrence texture features are selected for recognition. To improve the accuracy of the recognition, threshold filter, erosion filter and dilation filters are carried out after the SVM classification. Classification accuracy is measured by Kappa coefficients. The experimental results show that there is a well-recognized outcome. And it shows that texture-related features such as co-occurrence matrices might be high effective discriminators for high resolution remote sensing images and that SVM classification systems might lead to the successful discrimination of targets when fed with appropriate information.