Abstract
In multi-channel radar systems, accurate estimation of the clutter covariance matrix (CCM) is crucial for enhancing target detection and parameter measurement precision. However, the heterogeneous clutter poses significant challenges to CCM estimation, thereby degrading the performance of target detection and parameter estimation. To overcome this limitation, we propose a novel multi-channel radar detection method that exploits manifold projection. First, to enhance the suppression of strong clutter, the CCM estimation problem is reformulated on the Hermitian positive-definite (HPD) manifold, wherein the distributional characteristics of space-time coupled clutter samples are comprehensively analyzed. Then, based on the observed distribution of reference samples on the HPD manifold, we propose a sample transformation method designed for CCM estimation in the context of target detection. For this method, we derive a closed-form geodesic-distance lower bound (GDLB) on the HPD manifold, which explicitly quantifies the performance upper bound achievable by the proposed transformation. The proposed method weakens the heterogeneity and improves CCM accuracy by minimizing the geodesic distance among reference samples on the HPD manifold. Because of the curved, non- Euclidean geometry of the HPD manifold and the additional constraint imposed by the GDLB, the optimization problem is mathematically challenging. Therefore, we develop a manifoldprojection- based optimization algorithm and a parameter calculation method to ensure the optimization problem can be solved. By effectively transforming heterogeneous samples, the proposed method markedly enhances the target detection and parameter estimation performance of multi-channel radar systems. Experimental results obtained from both simulated and measured datasets validate the effectiveness and superiority of the proposed approach.
Funding Information
  • National Natural Science Foundation of China (62322122)
  • National Natural Science Foundation of China (62501616)
  • National Natural Science Foundation of China (62501605)
  • National Natural Science Foundation of China (62371458)

This publication has 66 references indexed in Scilit: