Detection Of Vehicle Targets In SAR Images

Synthetic Aperture Radar (SAR) has been widely applied to aerospace, ground reconnaissance and precision-guided weapon etc. SAR image target detection has become one of the core technologies in military applications. Cluster tanks, transport motorcade, the missile launchers are the important targets in the military battlefield. It has great significance and urgency to detect the tagets and to provide useful information for battlefield command. The ever-changing battlefield situations desire excellent real-time and effective detections, so fast-computing target detection algorithms of high-detection-rate and good robustness become the current research focus. In this thesis, according to the National Defence Prereserach Project, on the bases of studying fast SAR image speckle reduction, real-time CFAR target detection and road object extraction, we mainly research on automatic vehicle target detection of SAR images which adapts to complex scenario. The main content of this thesis are summarized as follows:1. We research on the speckle noise model and statistical distributions of SAR images, and propose a fast SAR image speckle reduction method base on complex derivative filters by the reference of sub-regional processing idea of EnLee filter algorithm. The canonical rotation and fast finite difference realization make excellent real-time processing nature of the algorithm. And With the inhibition of Gibbs phenomenon, we greatly improve the speckle reduction performance.2. We research on the Double-Parameter CFAR detection and the G0-distribution based CFAR detection algorithm, and propose a real-time SAR image CFAR detection methods by synthesizing the two algorithms. We gain the possible targets by giving the target confidence level and cumulating the image intensity. Then apply the DP-CFAR detection and G0-CFAR detection to the possible target pixels respectively. It not only greatly reduces the detection time, but also reduces the false alarm rate of detection.3. We analyze the road model of SAR images, and research on two road extraction algorithms based on the road model detection operator and complex derivative filter each. We combine the two results and extracte the road area by joining the broken edges and removing the false edges using the Hough transform. Lower-resolution processing reduces the amount of data dealing with, and the Simulation results meet the real-time requirements of detection system.4. We analyze the effects of target probability to target detection algorithm, and propose a real-time automatic vehicle target detection strategy which adapts to complex scenario by giving the target probability factor according to the road information and cluster information of military vehicles. And the results verify the effectiveness of the algorithm.