基于ES-ELM和FRFT的高频地波雷达多目标自适应检测

Adaptive detection of multi-targets of HFSWR based on ES-ELM and FRFT

高频地波雷达(high frequency surface wave radar,HFSWR)对于海事监测具有重要的军用及民用意义,然而在HFSWR回波信号中,待检测的目标常常淹没在海杂波和各种背景噪声中.因此,如何有效抑制杂波并实现多目标的自适应检测是HFSWR实现海事检测的关键和难点.该文提出了一种结合误差自校正极限学习机(error self-adjustment extreme learning machine,ES-ELM)和分数阶傅里叶变换(fractional Fourier transform,FRFT)的多目标自适应检测算法.算法根据相空间重构理论获得ELM的最佳状态空间,利用ES-ELM建立海杂波预测模型并对海杂波进行有效抑制;再在分数域根据目标信号的峰值集聚特征,利用Haar-like算子提取目标点的形态特征,并通过ES-ELM神经网络对目标进行自动辨识.实验结果表明,该文提出的算法具有良好的海杂波抑制能力,并可以实现海杂波背景下多运动目标的自适应高精度检测.

High frequency surface wave radar(HFSWR)has important military and civilian significance for maritime surveillance.However,in echo signal of HFSWR,the signals of targets to be detected are often submerged in sea clutter and various background noises.Therefore,effective suppression of sea clutter and adaptive detection of multi-targets are the key and difficult technique of maritime surveillance.In this paper,an adaptive detection algorithm of multi-targets combining error self-adjustment extreme learning machine(ES-ELM)and fractional Fourier transform(FRFT)is proposed.In the algorithm,phase space reconstruction theory is employed to obtain the optimal state space of ELM,and an ES-ELM is used to model and predict the sea clutter and suppress it effectively.In addition,according to the agglomeration characteristic of the peaks of objects in FRFT domain,Haar-like descriptor is employed to obtain the morphological features of target points in FRFT domain,and a proposed ES-ELM based method is used to identify the multi-targets automatically and adaptively.The experimental results show that the proposed algorithm has good sea clutter suppression ability and realizes adaptive detection of multi-targets in background of high sea clutter with high detection rate.