| Synthetic Aperture Radar (SAR) has now been widely used in military target detection and activecivilian radar sensing areas. Study on target detection technique and breakthrough in SAR picturesemi-automatic and automatic translation technique has been an urgent need for the development ofstrategic reconnaissance, monitoring, early warning and precision beating weapon.. In recent years,SAR has been put into applications in many areas and SAR can be used in target of interest detectionamong clutter background. In this dissertation, the moving and stationary target acquisition andrecognition, MSTAR database and Sandia database has been thorougly studied.First, the development of statistic characteristic of SAR image and target detection both at homeand abroad has been analyzed and comprehended. The problem arises that how to detect targetsamong complex background in big images.Then, two databases have been used in SAR image statistic characteristic acquisition. On thebases of K-S and KL index we can find how to affectively acquire the statistical model in differentbackground. Different backgrounds have been given for different distributions, including Gamma,Rician, K, Log Normal, G0and Weibull distributions. Also the limitations are put forward.Furthermore, CFAR detection has been investigated on the basis of statistic theory. The affect ofm resolution and model mismatch on detection performance has been mainly discussed and thethreshold values are deduced for various statistic models. The performance and charatersitics ofseveral basic detectors are analyzed.At last, in order to acquire complete and reliable target information for large image in complexterrian and avoid miss rate, we put forward a compress perception based target detection algorithm.Compared with canon wavelet algorithm, better index are abtained. By putting it into application ofSAR-CFAR for target detection, we can effectively avoid miss detection and the target completenessacquire5%furtherly. |
PDF Full Text Request