An Effective Multilevel Approach To Detect Changes In Buildings From Remote Sensing Images

Abstract

Change detection in satellite images taken over the same place at two different times is of primary importance in areas like disaster assessment, population growth estimation, land use monitoring, medical diagnosis and treatment, etc.. The exact location, accurate identification of objects and the extraction of appropriate image features form the basis for change detection in the images of interest. Locating the damaged buildings aids in effectively assessing the rate of damage assessment. In this work, various features such as geometric features, spectral band features and similarity metrics are exploited for efficient damage calculation. This work is mainly composed of five stages: preprocessing, FCM based clustering, Edge based segmentation and Feature extraction for both the pre-event image and the post-event image. The geometric features extracted from each segment of both the pre-event and post-event images are formed into a feature vector for further processing to detect the changes. The damage level index calculated using the similarity metrics was found to be 0.9737 % and accordingly, the post-event image was categorized under Very Slight Damage (VSD).