Research On Wideband Radar Target Detection Method Based On Multi-rank Subspace

Target detection is an important part of signal processing research and the basis for target tracking,imaging and recognition.Compared with narrowband radars,wideband radars have the advantages of high range resolution,strong anti-clutter capability,and low interception probability.Therefore,the use of wideband radars for target detection has important military and civilian values.But at the same time,they face the problems of non-negligible range migration and enhanced non-Gaussian clutter,so that the echo model and detection technology suitable for narrowband radar are no longer applicable,so how to design a detector based on the new model to improve the detection performance has become important research direction.In most of the existing work on wideband radar target detection,range spread target return signal is modeled as a group of vectors in the same rank-1 subspace but with different amplitudes.There are also some studies based on multi-rank subspace,but the above work only focuses on the effect of target return spread in several range cells caused by large bandwidth,while ignoring the range migration of the envelope across array element or during multiple pulses.In this paper,the adaptive detection of wideband radar targets in partially-homogeneous and non-homogeneous clutter environments is studied.A return signal model considering range migration is established,and the signal to be detected is modeled in the subspace formed by the signal matrix through appropriate transformation.Based on the model,the detection statistics are derived and the performance of the detector is analyzed.The content mainly includes the following three parts:First,giving a brief introduction to several traditional clutter models,then considering the pulse coherence and the high peak heavy tailing characteristics of the wideband radar clutter amplitude PDF,the Compound-Gaussian model widely used to describe non-Gaussian clutter is studied in detail.The simulation results show that the K distribution and Pareto distribution based on the model can fit the wideband radar clutter well.The processes of establishing the target return signal model and transforming it into the subspace are given,and then,the detection problem is deduced into a general binary hypothesis form,which lays the foundation for the design of detectors.Secondly,the match subspace detection theory and adaptive subspace detection theory are studied.Based on the return signal model,the subspace detectors are derived under conditions of complex Gaussian white and color noise and noise plus interference,and the detector characteristics are analyzed.Combined with the subband interpretation of the return signal model,subband form of subspace detectors with known and unknown covariance matrices under homogeneous environment and known and unknown noise power are given.In addition,subspace detectors under partially-homogeneous environment are also designed.The performance of the detector was studied by computer simulation,and the results showed that the performance of the detector we designed was very similar to the theoretical performance.Finally,in the compound-Gaussian clutter environment,considering Gamma and inverse Gamma random texture components,subspace detectors under non-homogeneous clutter environment are derived based on the proposed detection problem.Assuming that the clutter covariance matrix is known,the detection statistics are given and the MLE of the coefficient vector is calculated.Under the condition of sufficient secondary data,subspace detectors based on the covariance matrix RE estimation are derived.In the case of limited samples,the unknown parameters are all random variables with a given PDF,and assume that the covariance matrix follows the inverse Wishart distribution.The prior information of clutter is used to design subspace detectors based on the covariance matrix MAP estimation.The simulation of the performance of detectors show that our designed detectors have good performance when the target length is long,the clutter non-homogeneity is strong,and the prior information is mismatched.