基于微多普勒特征的雷达目标分类技术已被公认为最有发展潜力的技术之一，是当前雷达信号处理领域的研究热点，实现无人机载雷达对地海目标精确分类对于实现国防科技现代化具有战略意义。现有微多普勒效应研究均基于地基雷达、微多普勒特征提取过程复杂且不具备自适应性。本项目在考虑恶劣环境下无人机晃动、旋翼旋转和海面波浪涌动基础上，首次建立无人机载雷达对地海目标回波模型，分析其微多普勒特性。利用时频滤波实现地面目标微多普勒信号分离，利用奇异分解与重构实现海上目标微多普勒信号分离。基于双谱估计和改进的经验模态分解，实现不同目标的高区分度微多普勒特征快速提取和目标准确分类。最后，根据雷达电磁散射和微多普勒调制，构建和优化深度卷积神经网络，避免特征提取过程，实现地海目标智能分类。本项目的研究将实现无人机平台下，地海目标的精确分类，对基于微多普勒效应的雷达目标智能分类具有重要理论和应用价值。

Classification of radar targets based on micro-Doppler features has been recognized as one of the most promising technologies, which has been a research hotspot in the field of radar signal processing. Precise classification of ground and sea targets via (UAV) is of strategic significance for realizing the modernization of national defense. Existing studies are all based on ground radar and the micro-Doppler feature extraction process is complex and not adaptive. This project is the first to establish models of UAV-to-ground and UAV-to-sea targets and analyze their micro-Doppler characteristic after considering the drone sway, rotor rotation and sea wave surge in harsh environments. Then, time-frequency filtering is used to achieve the separation of micro-Doppler signal of ground targets and singular value decomposition is used to achieve the separation of micro-Doppler signal of sea targets. Based on bispectrum estimation and improved empirical mode decomposition, the high-resolution micro-Doppler features of different targets are quickly extracted and corresponding targets are accurately classified. Finally, according to radar electromagnetic scattering and micro-Doppler modulation, a deep convolutional neural network is constructed and optimized to avoid the feature extraction process and realize the intelligent classification of ground and sea targets. Research of this project will realize the precise classification of ground and sea targets using the UAV platform, having important theoretical and practical value for the intelligent classification of radar targets based on the micro-Doppler effect.