| 引用本文: | 王佳豪， 陈澍元， 赵书敏， 蒋忠进. 基于自适应阈值卷积网络的抗干扰雷达目标识别[J]. 雷达科学与技术, 2024, 22(5): 487-494.[点击复制] |
| WANG Jiahao, CHEN Shuyuan, ZHAO Shumin, JIANG Zhongjin. Target Recognition for Anti⁃Interference Radar Based on Adaptive Threshold Convolution Network[J]. Radar Science and Technology, 2024, 22(5): 487-494.[点击复制] |
|
| 摘要: |
| 本文提出了一种自适应阈值卷积网络（ATCN），基于HRRP数据进行抗干扰雷达目标识别。ATCN中的核心模块是自适应阈值卷积单元（ATCU），该模块能准确高效地完成对HRRP数据的特征提取。在ATCU中，采用自适应阈值函数充当激活函数，自动调整阈值以面对不同信干比的数据；利用多个不同尺度的卷积核来捕获HRRP数据中的区域差异特征；引入通道注意力机制和残差连接优化网络结构。本文进行了大量的抗干扰目标识别实验，实验结果表明，相比于所选择的3种对比网络，本文的ATCN网络能在不同干扰类型和不同信干比下提供更优的平均识别率和更好的指标稳定性，且具有更少的网络模型参数量和浮点运算次数，具备轻量化和高效的特点。 |
| 关键词: 雷达自动目标识别 高分辨距离像 压制性干扰 自适应阈值卷积单元 |
| DOI：DOI:10.3969/j.issn.1672-2337.2024.05.003 |
| 分类号:TN957.51 |
| 基金项目:国家自然科学基金资助项目（No.61890544，91748106） |
|
| Target Recognition for Anti⁃Interference Radar Based on Adaptive Threshold Convolution Network |
|
WANG Jiahao, CHEN Shuyuan, ZHAO Shumin, JIANG Zhongjin
|
|
1. State Key Laboratory of Millimeter Waves, Southeast University, Nanjing 210096, China;2. National Key Laboratory of Air?based Information Perception and Fusion, China Airborne Missile Academy, Luoyang 471009, China
|
| Abstract: |
| This paper proposes an adaptive threshold convolutional network （ATCN） for anti?interference radar target recognition based on HRRP data. The core module in ATCN is the adaptive threshold convolutional unit （ATCU）, which enables the accurate and efficient feature extraction from HRRP data. In ATCU, an adaptive threshold function is employed as the activation function to automatically adjust the threshold for different signal?to?interference ratios. Multiple convolutional kernels of different scales are used to capture regional difference features in HRRP data. The channel attention mechanism and residual connection are introduced to optimize the network structure. Extensive experiments on anti?interference target recognition are conducted in this study. The experimental results demonstrate that compared with the three selected comparison networks, the proposed ATCN provides better average recognition rate and better index stability under different interference types and signal?to?interference ratios. Furthermore, the ATCN network has fewer model parameters and floating?point operations, demonstrating its lightweight and efficient characteristics. |
| Key words: radar automatic target recognition （RATR） high resolution range profile （HRRP） suppression interference adaptive threshold convolution unit （ATCU） |
