Research On Knowledge-aided Clutter Suppression And Target Detection Algorithm For Airborne Radar

Space-time adaptive processing(STAP)is an effective technique to suppress ground clutter and improve the performance of moving target detection for airborne radar.The clutter covariance matrix estimation is the key factor to determine the capability of air-borne radar STAP.Whereas,the traditional STAP algorithms generally face the problem of small samples in heterogeneous clutter environment,which seriously affect the estimation accuracy of the covariance matrix,resulting in the performance degradation of clutter sup-pression and moving target detection.Through prior knowledge including the low-rank feature,structure of clutter covariance matrix,and etc,to eliminate redundant information in the echo signal and reduce the number of unknown parameters,it is a promising method that can significantly promote the estimation accuracy of the covariance matrix and the unknown parameters.Furthermore,with respect to the knowledge-aided STAP method,this dissertation also studies how to effectively improve the adaptability of airborne radar to complex environments and enhance the performance of moving target detection.This dissertation pursues the related researches in terms of the contents mentioned above,the main work and contributions are summarized as follows:(1)Knowledge-aided training sample selection algorithm for airborne radarAiming at the target self-nulling phenomenon caused by the contaminated samples,two training sample selection algorithms based on target and clutter knowledge are pro-posed respectively,which effectively improve the robustness of training sample selection.The proposed algorithms use prior knowledge to represent the property of the cell under test(CUT)clutter directly,through clutter spectrum reconstruction and the second-order spectral structure of the clutter covariance matrix.Thus the performance of CUT repre-sentation is not limited by the number of training samples and affected by the interference target signals.More importantly,the proposed two knowledge-aided methods take the spectral radius of the whitened sample covariance matrix and the difference of clutter spectrum energy as test statistics,which significantly improves the efficiency of sample selection.(2)Knowledge-aided clutter covariance matrix estimation algorithm for airborne phased array radarFocusing on the off grid problem of discrete dictionary based approach in the case of small samples,this dissertation proposes a gridless clutter covariance matrix estimation algorithm based on truncated nuclear norm for airborne phased array radar.By consid-ering the low-rank characteristics and the Block-Toeplitz structure as prior knowledge,the proposed method can estimate the clutter covariance matrix in the continuous domain,which solves the gridless problem caused by discrete dictionary.On the basis,a modified method with the truncated nuclear norm for clutter rank approximate is proposed.Com-pared with nuclear norm based approach,the modified method obtains higher accuracy and guarantees the convergence of the estimation.(3)Knowledge-aided clutter covariance matrix estimation algorithm for airborne multiple input multiple output(MIMO)radarTo improve the estimation accuracy of clutter covariance matrix for airborne MIMO radar under small sample scenarios,this dissertation presents a multiple structure knowl-edge based method.Specifically,the proposed method takes full advantage of the dou-ble Kronecker product structure of the MIMO radar in transmitting,receiving spatial and temporal domains,and fuses the clutter covariance matrix second-order spectral structure.The multiple structure knowledge effectively improves the estimation accuracy of the clut-ter covariance matrix with small samples.In addition,a proximal gradient algorithm is presented to solve the covariance matrix estimation of multiple structure knowledge con-straints,which realizes efficiently converge for the estimation process of the clutter co-variance matrix.(4)Knowledge-aided moving target detection algorithm based on direct data domainIn view of the performance degradation of moving target detection caused by insuffi-cient training samples in heterogeneous clutter environments,this dissertation develops a knowledge-aided detector based on direct data domain,aiming to effectively improve the moving target detection performance without samples scenarios.Concretely,taking the clutter subspace with redundant basis vectors and the structural feature of clutter covari-ance matrix as prior knowledge,the proposed method estimates the unknown parameters of CUT directly.Therefore,the detection performance is not affected by the heteroge-neous environment and clutter distribution,improving the ability to detect moving target in complex environment.