基于优化长短时记忆网络的海面微弱目标检测  

Weak Target Detection Based on Optimized Long Short-term Memory Network in Sea Clutter Background

作　　者：叶如 行鸿彦[1] 周星 YE Ru;XING Hongyan;ZHOU Xing(School of Electronic&Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学电子与信息工程学院,江苏南京210044

出　　处：《探测与控制学报》2024年第5期57-63,70,共8页Journal of Detection & Control

基　　金：国家自然科学基金项目(62171228);国家重点研发计划(2021YFE0105500)。

摘　　要：针对强混沌背景噪声下传统方法难以检测微弱目标信号的问题,研究了混沌相空间重构理论和麻雀寻优算法,提出一种基于优化长短时记忆网络(LSTM)的混沌背景下微弱信号检测方法。利用麻雀搜索算法优化LSTM模型参数,提高模型预测精度,降低目标检测门限,结合LSTM模型进行单步预测,利用预测误差从强海杂波背景下检测出微弱目标信号。以Lorenz混沌系统作为混沌背景进行仿真实验,对叠加的小信号进行检测,结果表明,该方法能够有效地检测微弱信号,其预测的均方根误差0.00171(信噪比为-137.707 dB),相较于传统神经网络预测模型、LSTM预测模型、GA-LSTM预测模型、PSO-LSTM预测模型均有显著提升。利用IPIX雷达信号进行预测实验,进一步验证了该方法的有效性。Aiming at the problem that traditional methods are difficult to detect weak target signals under strong chaotic background noise,this paper studied the chaotic phase space reconstruction theory and sparrow optimization algorithm,and proposed a weak signal detection method in chaotic background based on optimized long short-term memory network(LSTM).The sparrow search algorithm was used to optimize the parameters of the LSTM model,improved the prediction accuracy of the model,reduced the target detection threshold,combined the LSTM model for single-step prediction,and used the prediction error to detect weak target signals from the background of strong sea clutter.Using the Lorenz chaotic system as the chaotic background for simulation experiments,the superimposed small signals were detected,and the results showed that the proposed method could effectively detect weak signals.The predicted RMSE error of 0.00171(SNR=-137.707 dB)was significantly improved compared with the RMSE predicted by the LSTM model,and compared with the RMSE predicted by traditional neural networks.The prediction experiment using IPIX radar signal further verified the effectiveness of the proposed method.

关 键 词：微弱信号检测 长短时记忆网络 麻雀寻优算法 海杂波 

分 类 号：TN911.7[电子电信—通信与信息系统]