| 摘要: |
| 非相干检测器具有结构简单、计算复杂度低的优点，可满足雷达对海面中大型目标的实时检测需求。但实际探测环境中不可避免存在少量强回波野点，严重影响检测性能。为此，本文提出一种基于截断最大似然的恒虚警(Trimmed-Maximum-likelihood Constant False Alarm Rate, TML-CFAR)检测器，匹配于伽马分布海杂波的形状参数，并具有抗异常样本特性。首先，通过截断因子对参考单元数据进行选取，去除回波最弱和最强的异常样本，从而获得与待检测单元具有独立同分布的纯杂波。其次，在最大似然估计纯杂波功率的基础上，给出TML-CFAR检测器的检测统计量，推广到多脉冲伽马分布。然后，证明TML-CFAR检测器对于杂波尺度参数的恒虚警特性。同时，提出一种双分位点估计器，实现形状参数的稳健估计。最后，通过仿真数据和实测数据验证TML-CFAR检测器在不同杂波环境下的稳健性能，特别是在异常杂波下具有最佳性能。 |
| 关键词: 海杂波 目标检测 伽马分布 恒虚警 最大似然 |
| DOI： |
| 分类号:TN 959.1 |
| 基金项目:国家自然科学基金（62201184）； |
|
| Sea-surface Target Detection Based on Constant False Alarm Rate with Trimmed Maximum likelihood |
|
|
|
|
| Abstract: |
| Due to the advantages of simple structure and low computational complexity, non-coherent detectors can meet the real-time detection requirements of radars for sea-surface large targets. However, in the real environments, there are inevitably a small number of outliers with high power, which seriously affects the detection performance. Therefore, a trimmed maximum likelihood constant false alarm rate (TML-CFAR) detector is proposed in this paper, matching the shape parameters of gamma-distributed sea clutter and attaining the robustness to abnormal samples. Firstly, reference cells are selected by truncation factor to remove the weakest and strongest abnormal samples, thereby obtaining pure clutter with independent and identically distributed characteristics as cell under test (CUT). Secondly, under the condition of maximum likelihood estimation of pure clutter average clutter, the detection statistic of TML-CFAR detector is given and extended to multi-pulse Gamma distribution. Thirdly, it is proved that TML-CFAR detector has constant false alarm rate property with respect to the scale parameters of gamma clutter. Meanwhile, a bi-percentile estimator is proposed to achieve robust estimation of shape parameters. Finally, by using simulated data and real data, it is verified the robustness of TML-CFAR detector in different clutter environments, especially in the abnormal clutter where it exhibits the best performance. |
| Key words: sea clutter target detection gamma distribution constant false alarm rate maximum likelihood |