Modeling of change detection in VHF- and UHF-band SAR

Inhalt:
This paper describes models which can be used to evaluate change detection algorithms based on multiple SAR images in the VHF- and UHF-band. The models are evaluated using a processed SAR data set acquired with the airborne radar systems CARABAS-II (20-85 MHz, HH-polarization, 58deg incidence angle) and LORA (225-470 MHz, HH-polarisation, 50deg incidence angle) over a comparable deployment of ground vehicles concealed in forest. Five different models are investigated, i.e. Gaussian, complex Gaussian, log-normal, K-distribution and complex Gaussian + lognormal scene. It is shown that the latter model gives the best fit to both CARABAS-II and LORA data. The models are also used to simulate SAR images and evaluate the performance of four different change detection algorithms, i.e. amplitude ratio, Generalized Likelihood Ratio Test (GLRT) using amplitude data, GLRT using the logarithm of amplitude data and finally GLRT using complex data. The results indicate that different algorithms are optimal for CARABAS-II and LORA data, which is in agreement with results obtained from evaluations on real VHF and UHF radar data. It is shown that the reason is that the SAR data from CARABAS- II and LORA, respectively, have different statistical characteristics.