车载平台毫米波雷达三维点云视频成像  

3D Point Cloud Video Imagery of Automotive Millimeter-wave Radar

作　　者：崔硕 蒋梦杰 张邦杰 陈宇智 徐刚[1] CUI Shuo;JIANG Mengjie;ZHANG Bangjie;CHEN Yuzhi;XU Gang(State Key Laboratory of Millimeter Waves,Southeast University,Nanjing Jiangsu 210096,China)

机构地区：[1]东南大学毫米波全国重点实验室,江苏南京210096

出　　处：《现代雷达》2024年第10期32-38,共7页Modern Radar

基　　金：中央高校基本科研业务费资助项目(2242022K60008)。

摘　　要：针对毫米波雷达由于阵列分辨率限制等因素,其点云成像存在点云密度稀疏、精度低等问题,提出了一种基于合成孔径雷达(SAR)技术的车载平台毫米波雷达三维(距离、方位、俯仰)高分辨点云视频成像方法。首先,利用时域后向投影(BP)成像算法解决近距宽角域成像聚焦难题,获得高分辨二维视频SAR图像;然后,通过基于幅度阈值的复交替方向乘子法压缩感知网络(CV-ADMM-CSNet)获得场景高程信息,通过模型化方法并结合数据训练,实现快速实时高分辨三维成像;最后,结合多帧视频成像处理获得动态三维高分辨点云图像。仿真和实测数据实验,验证了本文算法的有效性。Due to the limitations of array resolution and etc,the point cloud imaging technology of millimeter-wave radar has the disadvantages of sparse distribution of point cloud and low-precision location.In this paper,a novel point cloud video imaging algorithm of 3D millimeter-wave radar on vehicle platform using the synthetic aperture radar(SAR)technology is proposed.First,the time-domain back-projection(BP)algorithm is applied to solve the slant-range wide-angle imaging focusing problem,obtaining well-focused high-resolution 2D image.Second,the amplitude threshold-based Complex-Valued Alternating Direction Method of Multipliers Compressed Sensing Net(CV-ADMM-CSNet)is applied to obtain elevation information of scenes,which achieves fast and real-time high-resolution 3D imaging through modeling methods combined with data training.Finally,the dynamic 3D point cloud image can be successfully generated in combination with the multi-frame SAR imaging.The experimental analysis using simulated and measured data is performed to confirm the effectiveness of the proposed algorithm.

关 键 词：毫米波雷达 三维点云视频成像 合成孔径雷达 

分 类 号：TN957.52[电子电信—信号与信息处理]