Hide/Show Apps

New change detection method using double segmentation and its application on remotely sensed images

Gedik, Ekin
Change detection research, a branch of statistical data analysis, focuses on detecting changed samples between di erent observations of the same dataset. The proposed study presents a novel change detection procedure and its application as a complete framework which is designed to work on remotely sensed images. The scope of the study is defined as detecting man-made change objects between satellite images of the same region, acquired at di erent times. Proposed framework has three main steps as preprocessing, feature extractionclassification and postprocessing. Preprocessing step normalizes, registers and measures the similarity of image pairs. The main contribution of the proposed study lies at the feature extraction and classification step. With the help of newly proposed "double segmentation" paradigm, an object based approach can be utilized without any prior information or supervision. Well known features in the change detection literature are defined, combined and compared in the study. Apart from known classification methods such as K-Means Clustering and Expectation-Maximization, a novel heuristic thresholding method is also presented. A postprocessing procedure which helps to obtain more accurate and visually appealing results is also provided. Experiments conducted on artificial and real satellite images show that proposed framework is good at capturing the man-made change object characteristics in remotely sensed images with high accuracy.