Skip to Main Content
Change detection using ultra-wideband synthetic aperture radar (SAR) images in the low end of the VHF band is shown to provide excellent performance for detection of vehicle-sized objects in forest concealment. Two different change detection algorithms are discussed and their performance evaluated. The two algorithms are based on similar statistical hypothesis testing, but differ in that one operates on complex (coherent change detection) whereas the other uses magnitude (incoherent change detection) image data. Algorithm evaluation is performed using radar data acquired with the airborne CARABAS-II SAR in northern Sweden. The data were collected during a change detection experiment with concealed vehicles in boreal forests (stand volume ca. 100 m3/ha). Results show that coherent change detection gives slightly better performance using full spatial resolution of the images, whereas the incoherent change detection gives better performance when spatial averaging (2×2 resolution cells) is included. A comparison with detecting vehicles using only single-pass images shows an increase of false alarms of one to two orders of magnitude at the same probability of detection.