Research On Clutter Suppression Method Based On Knowledge Aided For Airborne Radar

When the airborne radar detects moving targets,it will receive echo that contain a lot of powerful ground clutter,which will reduce the target detection performance of the radar system.STAP technology combined with time-domain and space-domain sampled data for two-dimensional adaptive filter can suppress ground clutter effectively.Therefore,STAP technology has become a key technology for airborne radar clutter suppression.STAP technology usually assumes that the system can obtain enough identically distributed and independent training samples to estimate the clutter covariance matrix.However,in practice,clutter is often heterogeneous.This situation will cause the estimated clutter covariance matrix to be inaccurate.The clutter suppression performance of the STAP algorithm is reduced.Therefore,the new STAP algorithm applicable in heterogeneous clutter environment has become the research direction of STAP technology.Reasonable use of prior knowledge in the environment to achieve knowledge-aided space-time adaptive processing(KA-STAP)has become a research hotspot for improving the performance of STAP algorithms in heterogeneous clutter environments.In this paper,the KA-STAP technology under the heterogeneous clutter environments is studied,mainly to study two methods to improve the estimation accuracy of covariance matrix.The main content of this paper is summarized as follows:1.The method of constructing a priori covariance matrix of samples under heterogeneous clutter environment to select training samples and weight training samples is studied.When the radar exposure scene is severely heterogeneous,the sampling covariance matrix used to estimate the real clutter covariance matrix will produce larger errors.The generalized inner product(GIP)algorithm can excise the heterogeneous training samples from initial training sample set,but the algorithm doesn't take the property of the cell under test into consideration,that is,the selected samples may be homogeneous with each other,but are not consistent with the characteristics of the cell under test;In the process of solving the sampling covariance matrix,all training samples have the same weight,which is also unreliable.In response to these problems,a weighted KA-STAP algorithm is proposed,which uses terrain information and elevation information in the geographical environment to estimate the prior clutter covariance matrix of each distance unit and then selects training samples.And the weight of each training sample is derived according to the similarity between the prior covariance matrices,which can effectively overcome the differences between the training samples and more accurately estimate the clutter covariance matrix of the cell under test.Experimental results using the real data show that the clutter suppression capability of this method is improved by 2 ~ 2.5d B compared with the GIP algorithm.2.The covariance matrix estimation method for selecting samples directly by using the fine prior knowledge of the surface of the Range-Doppler unit is studied.The statistical characteristics of the clutter are inseparable from the coverage type of its location table.According to this fact,the KAT algorithm in the existing KA-STAP method calculates the proportion of each terrain in each RD unit and selects training samples according to the proportion.This method has achieved better clutter suppression effect.However,the KAT algorithm does not consider the relationship between the reflection characteristics of the terrain types;nor does it consider the influence of the specific distribution of the strong and weak points in the RD unit on the overall clutter characteristics of the unit.To solve these problems,a KA-STAP algorithm for RD unit azimuth fine division is proposed.First,each RD unit is segmented from Doppler,then the reflection coefficients of each part in the RD unit are obtained by using terrain data NLCD,and finally the reflection coefficient distribution of all range units is used to compare the similarity between the training samples and the cell under test,and select the training samples.Experimental results using the real data show that the clutter suppression capability of this method is improved by 2 ~ 3d B compared with the KAT algorithm.