Change Detection in Sar Images

A system has been developed which is able to detect changes in an area by using two satellite SAR images taken from similar tracks, but from different passes. Accurate registration is performed by correlation on a subpixel basis. The change detection then involves taking the ratios rather than differences. Different ratio masks are used to detect different features. Threshold values used to produce the final change detection image, is dependent on the ratio mask size. Thresholds are estimated by calculating the probabilty of false alarm (pfa) for different threshold values until pfa I 1 % for the ratio image. Tests have been carried out using two SEASAT SAR images.