现代防御技术 ›› 2021, Vol. 49 ›› Issue (6): 74-83.DOI: 10.3969/j.issn.1009-086x.2021.06.012

• 目标特性与探测跟踪技术 • 上一篇    下一篇

遗传算法优化神经网络的雷达杂波抑制方法

施端阳1,2, 林强1, 胡冰1, 陈嘉勋3   

  1. 1.空军预警学院,湖北 武汉 430019;
    2.中国人民解放军95174部队,湖北 武汉 430040;
    3.中国人民解放军95980部队,湖北 襄阳 441022
  • 收稿日期:2021-05-20 修回日期:2021-08-17 发布日期:2022-01-12
  • 作者简介:施端阳(1992-),男,湖北枝江人。博士生,主要从事预警装备发展论证研究。通信地址:430019 湖北省武汉市江岸区黄浦大街288号 E-mail:644481846@qq.com
  • 基金资助:
    全军军事类研究生资助课题(JY2020B150)

Radar Clutter Suppression Method Based on Neural Network Optimized by Genetic Algorithm

SHI Duan-yang1,2, LIN Qiang1, HU Bing1, CHEN Jia-xun3   

  1. 1. Air Force Early Warning Academy,Hubei Wuhan 430019,China;
    2. PLA,No. 95174 Troop, Hubei Wuhan 430040,China;
    3. PLA,No. 95980 Troop,Hubei Xiangyang 441022,China
  • Received:2021-05-20 Revised:2021-08-17 Published:2022-01-12

摘要: 针对目标检测后仍存在剩余杂波影响雷达跟踪和占用数据处理资源等问题,提出一种基于遗传算法(genetic algorithm,GA)优化反向传播(back propagation,BP)神经网络的雷达杂波抑制方法。通过分析雷达目标点迹和杂波点迹差异化的特征,选取多维特征作为输入自变量,设计BP神经网络分类器模型,对雷达目标点迹和杂波点迹进行分类识别,进而滤除杂波。在数据输入时,利用遗传算法对神经网络的输入自变量进行优化选择,降低输入数据维度,缩短建模时间。在神经网络训练时,利用遗传算法优化BP神经网络的初始权值和阈值,提高网络收敛速度和识别精度。通过雷达实测数据测试表明:遗传算法优化后的BP神经网络与传统BP神经网络相比,雷达杂波点迹识别率提高了1.5%,识别时间缩短了20.4%。

关键词: 遗传算法, 反向传播神经网络, 降维, 点迹数据, 剩余杂波, 杂波抑制

Abstract: Aiming at the problems that the radar tracking and occupies data processing resources are affected by the residual clutter after target detection,a radar clutter suppression method based on genetic algorithm (GA) optimized back propagation (BP) neural network is proposed.By analyzing the differentiated features of radar target plots and clutter plots,multi-dimensional features are selected as input independent variables,and BP neural network classifier model is designed to classify radar target plots and clutter plots,so as to filter clutter.During the data input, genetic algorithm is used to optimize the input independent variables of neural network to reduce the dimension of input data and shorten the modeling time.During the neural network training process,GA is used to optimize the initial weights and thresholds of BP neural network to improve the network convergence speed and recognition accuracy.Tests on radar measured data show that:the BP neural network optimized by genetic algorithm has a 1.5% increase in the recognition rate of radar clutter plots and a 20.4% reduction in recognition time compared with the traditional BP neural network.

Key words: genetic algorithm (GA), back propagation (BP) neural network, dimension reduction, plot data, residual clutter, clutter suppression

中图分类号: