基于改进ACGAN的雷达空中目标细分类方法  

Fine Classification Method of Radar Air Targets Based on Improved Auxiliary Generation Countermeasure Network

作　　者：刘帅康 曹伟 管志强 杨学岭 许金鑫 LIU Shuaikang;CAO Wei;GUAN Zhiqiang;YANG Xueling;XU Jinxin(Nanjing Shipborne Radar Research Institute,Nanjing 210000,China)

机构地区：[1]南京舰载雷达研究所,南京210000

出　　处：《火力与指挥控制》2023年第7期74-78,84,共6页Fire Control & Command Control

摘　　要：为了解决窄带雷达空中3类飞机目标难以细分类的问题,提出了一种基于改进辅助生成对抗网络(auxiliary classifier generate adversarial networks,ACGAN)方法,将卷积神经网络(convolutional neural networks,CNN)结合堆叠的双向长短期记忆网络(bidirectional long short-termmemory,Bi-LSTM)嵌入到ACGAN中,使ACGAN具有处理目标频域内部时序特征的能力。通过对X波段对空警戒雷达实测数据对比实验表明,提出的方法能够有效地对空中目标进行细分类,并具有较高的识别正确率。In order to solve the problem that it is difficult to subdivide the three types of aircraft targets in the narrowband radar in the air,a method based on the improved auxiliary classifier generate adversarial networks(ACGAN)is proposed,convolutional neural networks(CNN)and stacked Bidirectional Long Short Term Memoryy(Bi-LSTM)are embedded into ACGAN to make ACGAN have the ability to deal with the internal sequential characteristics of the target frequency domain.The comparison experiment of the measured data of the X-band air warning radar shows that the proposed method can effectively and finely classify the air targets and has a high recognition accuracy.

关 键 词：窄带雷达 空中目标分类 辅助生成对抗网络 双向长短期记忆网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]