基于对角积分双谱的海面慢速小目标检测方法

针对海杂波背景下雷达对海面慢速小目标探测技术难题，该文提出一种基于对角积分双谱的三特征融合检测方法。该方法首先从待检测信号的估计双谱中获得对角积分双谱，而后根据海杂波单元与目标单元之间的非线性耦合差异性，进一步从对角积分双谱中提取峰值、质心频率、谱宽3种特征。考虑到扫描模式下雷达采用的相干脉冲数通常较少，易导致特征不稳定，进而影响海杂波与目标可分性，为此，通过多帧扫描历史数据和当前帧数据的综合应用，对谱特征进行积累得到累积峰值、全变差、累积谱宽3种累积特征。最后采用凸包分类算法，在三特征空间进行融合检测。经实测CSIR数据集验证，在同等参数条件下，该文检测方法相比已有基于时频三特征的检测方法，基于幅度、多普勒三特征检测方法和分形特征检测方法具有更好的检测性能。

A Method for Detecting Small Slow Targets in Sea Surface Based on Diagonal Integrated Bispectrum

Considering the technical difficulty of radar to detect small targets embedded in the sea clutter, a three-feature fusion detection method based on diagonal integrated bispectrum is proposed. Firstly, the diagonal integrated bispectrum is obtained from the estimated bispectrum of the signal to be detected. Then, according to the nonlinear coupling difference between sea clutter cell and target cell, three features consist of peak value, centroid frequency and spectrum width are extracted from the diagonal integrated bispectrum. Considering that the number of coherent pulses used by radar in scanning mode is usually small, it is easy to lead to feature instability, and then affect the separability of sea clutter and target. For this reason, through the comprehensive application of multi-frame scanning historical data and current frame data, three cumulative features including cumulative peak value, total variation, cumulative spectrum width are obtained by accumulating three spectrum features. Finally, the convex hull classification algorithm is used to perform fusion detection in three dimensional feature space. The measured CSIR dataset verifies that, under same parameters, the proposed detection method has better detection performance compared with the existing detection methods based on three time-frequency features, amplitude feature and doppler features, fractal feature.