探测器
物理
子空间拓扑
接头(建筑物)
数学
算法
统计
矩阵分解
探测理论
概率密度函数
分解
信号处理
计算机科学
估计理论
信噪比(成像)
噪音(视频)
数学分析
滤波理论
矩阵代数
时频分析
雷达跟踪器
噪声测量
信号子空间
联合概率分布
作者
Jiachen Zhu,Fangjiong Chen,Jie Wu,Ming Yi Xia
标识
DOI:10.1109/taes.2026.3737087
摘要
Existing cell-by-cell constant false-alarm rate (CFAR) detectors are susceptible to locally coupled spectral spreading around the cell under test (CUT), where window-induced main-lobe broadening and sidelobe leakage of the CUT overlap with leakage from nearby scatterers within a small local patch. As a result, the reference window becomes contaminated and nonhomogeneous, biasing threshold setting and increasing false alarms and missed detections. Although existing iterative or sidelobe-suppression-based CFAR methods improve interference robustness, they either fail to explicitly restore reference-cell homogeneity or rely on a locally isolated-target assumption, thereby neglecting the multi-component coupling induced by neighboring targets. To address these issues, this article proposes TJC-CFAR, an iterative CLEAN-based CFAR detector that integrates truncated statistics with joint subspace decomposition to mitigate locally coupled spectral spreading. First, a windowed truncated-statistics noise estimator is developed to estimate the background interference level and improve reference-cell homogeneity for reliable threshold setting. Second, a joint subspace decomposition and multi-scale reconstruction module analyzes the CUT-centered local block to construct a coupled model with multiple components, enabling component-wise decoupling and accurate amplitude estimation at the CUT. In addition, a CLEAN-based iterative procedure is employed to suppress target-induced sidelobes and refine detections. Experiments on both simulated and measured data show that TJC-CFAR reduces false alarms, improves target-detection precision, and achieves superior overall detection performance compared with representative baselines detectors, while maintaining manageable computational complexity.
科研通智能强力驱动
Strongly Powered by AbleSci AI