High Resolution SAR Images Complex Ield’ Research And Implementation Of Hange Detection Method

Synthetic Aperture Radar (Synthetic Aperture Radar, SAR) is anmicrowave imaging radar which has strong penetrating power, and worksall day, SAR provides images of high resolution on ground objects.Interferometric SAR (Interfermetric SAR, InSAR) is a technology whichsets a SAR system on the satellite or aircraft to get multiple high-resolutioncomplex SAR image of the same areas. In the process of image acquisition,due to space baseline error, the effect of atmospheric changes, track error,system thermal noise, registration and other factors, sometimes thesefactors make the image’s amplitude and phase information changed, thenmake the change detection results have a high false alarm rate.This paper aimed to reduce the false alarm rate.In the process of registration, we proposed a method which uses thehigh-resolution SAR images’ geographic information to do registration,then combined with coarse and fine registration to finish other process ofregistration. This method improve the efficiency of registration and can getprecise results, in this process we save the computing time and storagespace. In part of experiments, we use two sets of data to verify thefeasibility of the method.In terms of change detection, based on the traditional correlationcoefficient estimator, we propose a more stable estimator and give changedetection method (Complex change detection) which make Fishertransformation to the estimator. we combine the simulated ROC curveswith the experiment results that we get from data to prove that theproposed method has better effect on the aspect of reducing false alarms. Considering the single detection method may not be accurate whendetecting scenes which have a lot of low coherence areas, we propose achange detection framework which combine the method of using images’amplitude to do change detection with the method of complex changedetection method which uses the complex images to do change detection.In part of experiment, we use images that coming from two differentscenarios to testing the framework we proposed. Experiments results showthat the framework we proposed have a more accurate change detectionresults for areas which exist a large number of low-coherence, then verifiesthat the feasibility of proposed change detection framework.