Road extraction and its application in moving target detecion in SAR images

This paper discusses a moving target detection method based on road extraction in SAR images. Firstly road pixels are initially extracted by fuzzy C mean (FCM) clustering. Then the image is divided into sub-block and the Hough transform is used to extract the segment candidates in each block. Combining the proximity, the con-linearity and the membership of the segment candidates, the reasonable roads are preserved and false roads are removed. Integration of a priori information on road traffic network, making use of binary statistical hypothesis testing and change detection method, moving target detection performance in SAR images is improved. Simulation results demonstrate the effectiveness of the method.