激光点云运动人体目标识别及运动信息提取  

Human target recognition and motion information extraction based on laser point cloud

作　　者：杜璞[1] 张小艳[2] 

机构地区：[1]陕西警官职业学院,西安710043 [2]西安科技大学,西安710054

出　　处：《激光杂志》2018年第1期150-153,共4页Laser Journal

基　　金：国家自然科学基金青年项目(No.F030504)

摘　　要：针对传统的关键点检测方法进行人体运动信息提取和图形重构获得的可测截面少和精度低的问题,提出一种基于激光点云技术的运动人体目标识别及运动信息提取方法。采用多视点云配准方法进行运动人体目标的三维光学面扫描,获得运动人体目标的体绘制图像,采用高斯滤波法剔除服从正态分布的噪声,获得能准确反映人体运动行为特征的有用数据特征。使用激光点云平滑技术消除人体运动重建曲面的毛刺,保留原始点云的特征信息。通过点云采样技术拟合获得人体运动信息,实现人体运动信息有效提取。仿真结果表明,该方法能有效实现人体运动动作图像重构,重建几何精度较高,对运动信息提取的拟合准确性较好。Aiming at the drawbacks of traditional detection methods that less measurable cross-section and low precision of human motion information extraction and graphics reconstruction, this paper presents an approach to ex- tract human motion recognition and motion information technology based on laser point cloud. Scanning 3 D optical sec- tion of moving human by multi- perspective registration method to obtain volume rendering image. Filter the noise which meet the normal distribution through Gauss filtering to obtain useful features which can accurately reflect human movement behavior. Use laser point cloud smoothing technology to eliminate the human motion reconstruction surface burr and retain original point cloud feature information. The human motion information is fitted by point cloud sampling technology to realize the effective extraction of human motion information. The simulation results show that this method can effectively reconstruct the human motion image and has high geometric precision.

关 键 词：激光点云 人体 目标识别 运动 信息提取 图像 

分 类 号：TN249[电子电信—物理电子学]