Urban area detection using gabor features and spatial voting

Abstract

Monitoring urbanization is an important problem in remote sensing. Very high resolution satellite images provide valuable information to solve this problem. However, these images are not sufficient alone for two main reasons. First, a human expert should analyze very large images. There may always be some errors in operation. Second, urban regions are dynamic. Therefore, monitoring urbanization should be done periodically. This is time consuming. To handle these shortcomings, an automated system is needed to monitor urbanization. In this study, we propose an automated method to detect urban areas in very high resolution satellite images. Our work can be taken as the first and most important step for automatically monitoring urbanization. Our method is based on local feature extraction using Gabor filters. These local features are represented in a voting space. Using it the urban regions are detected automatically. We test our method on a diverse panchromatic Ikonos image set. Our test results indicate the possible use of our method in practical applications.