LFMCW雷达人体目标探测与特征识别研究

发布时间：2018-01-18 14:31

本文关键词：LFMCW雷达人体目标探测与特征识别研究　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：由于人体目标雷达散射截面积较小,运动速度缓慢,其多普勒频率与地物杂波重叠,在强地物杂波环境中对慢速人体目标探测与识别一直是雷达面临的严峻挑战。本文采用线性调频连续波体制雷达针对人体探测与特征识别课题开展如下研究:首先,简化Boulic模型为12关节、10散射部分的人体运动模型,在人体坐标系下完善人体12关节旋转角度公式,10散射中心点位移公式,基于此构建LFMCW雷达人体运动回波模型。通过对比实测数据与仿真模型的时频谱图,验证LFMCW雷达人体运动回波模型的正确性。其次,针对LFMCW雷达采用长时间相干积累探测运动人体目标存在多普勒扩展问题,设计基于相位补偿变换的相干积累方法,该算法通过循环搜索寻找最佳人体运动参数,据此构建相位补偿信号匹配行进人体差拍信号中的非线性相位,在相干积累时有效利用多普勒扩散能量,提高检测信噪比,改善LFMCW雷达对人体目标的探测能力。仿真与实测数据分析结果验证了该算法的有效性。最后,针对LFMCW雷达人体运动识别问题,在时频域分析人体运动模型与实测数据在多种运动形式下的微多普勒变化规律,归纳人体运动物理特征值,据此设计支持向量机分类树结构的人体运动形式分类器。在大量实测数据的基础上,分析分类器采用不同人体运动特征值时的识别性能。
[Abstract]:Because the human body target RCS is smaller, speed is slow, the Doppler frequency and the ground clutter overlap in strong ground clutter environment to slow human target detection and recognition has been a severe challenge for radar. This paper adopts linear frequency modulated continuous wave radar system for human detection and recognition task to carry out the research as follows: first, a simplified Boulic model for 12 joint human motion model 10 scattering part of the rotation angle of the human body 12 joint formula in the body coordinate system perfect, 10 scattering center displacement formula, the construction of LFMCW radar echo model based on human motion. The spectrum of the measured data and simulation model, the correctness of model validation LFMCW the radar echo of human motion. Secondly, according to the LFMCW radar using long time coherent detection of moving human target Doppler expansion problem, based on the design phase A method of coherent accumulation compensation transformation, the algorithm iteratively searches for the best human motion parameters, based on phase compensation signal, pedestrian beat nonlinear phase signal, the effective use of Doppler diffusion energy in coherent accumulation, improve the signal-to-noise ratio and improve the detection ability of LFMCW radar for human target. The simulation and measurement the data analysis results show the effectiveness of the algorithm. Finally, LFMCW radar for human motion recognition problems in time-frequency analysis of human motion model and the measured data in a variety of sports in the form of micro Doppler changes, summed up the value of physical characteristic of human motion, based on support vector machine classifier form human motion classification tree structure in. Based on the measured data, analysis of different classifiers using human motion recognition performance value.

【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN957.51

【参考文献】