Research On Radar Target Detection Under Complex Clutter Background

Research on radar target detection technology in homogeneous and stationary clutter environment has become increasingly mature,however,the actual complex geographic location leads to heterogeneous and non-stationary characteristics of clutter background.It's no longer reasonable to model the clutter as a single Gaussian distribution,and the selected reference cells aren't subject to the same distribution,resulting in a siganificant decrease in detection performance.The thesis is under the background of the surveillance radar system.It classifies the clutter area based on the statistical characteristics difference of clutter,then identifies the clutter distribution model on the basis of clutter partition.After that,the background is uniformly converted to hypothesis model of the traditional detector by means of clutter model transformation,which solves the problem that heterogeneity and non-stationarity of clutter in practical environment lead to the increase of false alarm.Finally the detected clutter false alarm is suppressed by the plot processing algorithm.The specific contents of this thesis are as follows:1.In the case of complex detection background of the measured data,K-L divergence method is firstly used to partition the detection background.This algorithm is based on the clutter statistical distribution differences,and sets the threshold according to the calculated K-L distance to achieve the purpose of classify background.And then the threshold methods of image segmentation are studied.According to the traditional Ostu and the Ostu based on the variance information,a weighted Ostu is proposed to automatically determine the segmentation threshold.The K-L divergence and weighted Ostu are used to divide the detection background into three regions: homogeneous,weak clutter and strong clutter.Finally,the partition results are smoothed to eliminate the discontinuous and small areas in each region.2.The design of CFAR detector is usually based on the identified clutter model.In order to select the appropriate detection algorithm,the clutter statistical distribution of each region needs to be studied.Firstly,the commonly used clutter distribution models and parameter estimation methods are discussed.Two kinds of clutter test methods are used to quantitatively analyze the clutter distribution characteristics of each region of the measured data,and we achieve acquiring knowledge of the detection background.The prior information extracted layes the foundation for the following design of the detection strategies.3.Against the practical detection background where there are a variety of distribution types of clutter,this thesis proposes the strategy of partition and normalization,accomplishing the clutter distribution model conversion in each region and index normalization of detection background,to realize the background to obey the same parameters of the exponential distribution.The method achieves the purpose of improving detection performance by matching the detector model and homogenizing the background.Under the simulated and measured data,compared with the traditional CFAR and adaptive detection algorithms,the performance comparison results prove the superiority of the proposed method.Finally,the plot processing algorithm is studied to further suppress the false plots of detector.