CFAR Detection Of Targets In SAR Images Based On Statistical Model

Synthetic aperture radar (SAR) has a wide field of application with good prospectsin military reconnaissance. And detection and discrimination of military targets in aSAR image is an important task for acquirement of information from the battlefield. It isvery meaningful to extract the information of the attractive targets from a SAR imagequickly and locate the targets accurately.The aim of this article is to provide a systematic method to study the constant falsealarm rate(CFAR) algorithm for target detection in SAR images based on statisticalmodel. Then the paper studies the CFAR algorithm which is used for detecting targetsin SAR images based on mixture gaussian model. In order to overcome the shortcomingthat mixture gaussian model is not suitable for nonhomegeneous SAR images, the paperput forward a CFAR algorithm based on generalized gamma distribution model fordetecting targets in SAR images. The method estimates the parameters of thegeneralized gamma distribution in a SAR image using logarithmic cumulants. And theconcrete forms of CFAR algorithm for generalized gamma distribution is deduced. Thetargets and non-targets in a SAR image are distinguished by the CFAR threshold. Thenunconnected targets regions are cluster by class distance. And multi-feature fusionstrategy is adopetd to discriminate targets and non-targets so as to reduce false alarmrate in the SAR image. At last, the area of interesting targets will be marked.The experimental data is the real SAR images of Sandia National Laboratory andsimulated high-resolution SAR images based on the MSTAR database. According to thefitting histogram results, the method of generalize gamma distribution is better thanmixture gaussian considering that it is more accurate. The experimental results obtainedverify the effectiveness of the algorithm studied in this paper.