无人机产业的快速发展在带来无人机广泛应用的同时也造成了潜在的威胁，因此对无人机进行分类识别具有重要意义。雷达微多普勒效应以其能够反映目标局部运动属性的信息特征在雷达目标分类识别领域获得了广泛的研究应用。本文围绕提高无人机目标识别性能这一主题，研究了基于时频分析的雷达微多普勒效应无人机目标特性提取与识别方法，主要完成了三项工作：1）对三类无人机的数据回波数据预处理，利用时频变换获得无人机的微多普勒特性时频图。然后利用主成份分析的方法完成了无人机微多普勒特征的提取，降维后的数据特征极大降低了后续识别的运算量，通过KNN分类器进行无人机数据的训练与测试识别，最终得到优于87%的识别准确率。并且分析对比了这一方法与利用波形特征和频谱特征识别方法的性能。2）通过双雷达观测微动特征融合的方法，提升了无人机识别的准确率。通过搭建的多角度观测雷达传感器系统收集三类无人机的回波数据，采用时频变换加主成份分析的方法完成了无人机的微多普勒特征提取，并利用多角度雷达传感器观测的数据进行特性融合来代表无人机。通过SVM分类器进行无人机目标的训练与测试识别，最终识别准确率达到97.7%。此外分析了不同的训练样本比例以及不同的信噪比下对该方法性能的影响。3）研究了利用雷达多径信号的微多普勒特征提升无人机识别性能的方法。利用调频连续波雷达收集三类无人机的多径回波数据，通过主成分分析方法在无人机的时频图上获得微多普勒特征并进行多径信号的微多普勒特征融合来代表无人机，将微多普勒融合特性数据代入分类器进行训练测试识别，最终获得相对于仅用直达波信号处理提升平均5%的分类准确率。此外分析了不同的参数对该方法的性能的影响。

Recognition of drones is important due to their rapidly development and potential threats. Micro-Doppler-based methods for radar target recognition and classification has been widely studied for its accurate representation of the local motion characteristics of targets. This thesis is focused on the task of improving the performance of drones recognition based on micro-Doppler radar signatures. Three main tasks have been accomplished as follows:1) Time-frequency spectrograms are obtained by performing a short-time Fourier transform on the radar preprocessed echo data. Then, principal components analysis is utilized to extract the drones micro-Doppler features. The exact characteristic of the drones greatly reduce the computation of the subsequent classification recognition. The classification accuracy is better than 87%, by the training and testing based on KNN classifier. And also, the proposed method is analyzed and compared with other methods.2) Dual radar sensing classification scheme is proposed to enhance the robustness of micro-Doppler based classification of drones. Through the short-time Fourier transform and principal components analysis, the drones' micro-Doppler feature is extracted and the features obtained by the two radar sensors are fused together to representing the drones. The experimental results show that the classification accuracy is better than 97.7% based on the SVM classifier. In addition, the performance of the proposed method is also analysed with different sizes of training dataset and different noise levels. 3) Multipath micro-Doppler radar signatures is presented to improve the accuracy of micro-drone classification in urban environments. By using a high-resolution radar, the direct-path and multipath echoes are separated based on the indices of the range cells they occupy. The time-frequency spectrograms of the drones are obtained by performing the short-time Fourier transform on the direct-path and multipath echoes, respectively. The direct-path and multipath features, which are extracted by the principal components analysis on the direct-path and multipath spectrograms, respectively, are fused and fed into classifiers to determine the type of drones. Experimental results based on measured data demonstrate that the classification accuracy produced by multipath exploitation is 5% higher than that obtained by using the direct-path echo only. In addition, the performance of the proposed method is also analysed with different parameters.