针对传统的雷达动目标检测方法在杂波背景下目标识别率低的问题,提出了基于时频分析和卷积神经网络的雷达动目标检测方法。首先,通过同步提取变换将动目标的回波信号转换为时频分布,初步提取回波信号的时频特征;然后,对回波信号时频分布的脊线进行提取,并基于此构建数据集;最后,将数据集输入AlexNet进行训练和测试,实现雷达动目标的识别和分类。仿真实验表明,基于SET和AlexNet的方法在噪声环境下能够有效检测动目标,对匀速、匀减速、匀加速三类动目标都具有较高的识别率。脊线提取的应用增强了低信噪比下回波信号的时频特征,提高了检测方法的准确率和噪声鲁棒性。

Aiming at the low recognition rate of traditional radar moving target detection methods in clutter background, a radar moving target detection method is proposed based on time-frequency analysis and convolution neural network. Firstly, the echo signal of moving target is converted into time-frequency distribution through synchroextracting transform, and the time-frequency characteristics of the echo signal are initially extracted. Then, the ridge is extracted from the time-frequency distribution of the echo signal, and the data set is constructed by the ridge. Finally, the dataset is input into AlexNet for training and testing to realize the recognition and classification of radar moving targets. The simulation results show that the method based on SET and AlexNet can effectively detect moving targets in noisy environment, and has high recognition rate for three types of moving targets:uniform speed, uniform deceleration and uniform acceleration. The application of ridge extraction enhances the time-frequency characteristics of the echo signal under low signal-to-noise ratio, and improves the accuracy and noise robustness of the detection method.

TN959

国家自然科学基金资助项目(62201298, 11801287);内蒙古科技大学创新基金资助项目(2019QDLB39);内蒙古科技大学基本科研业务费专项资金