车窗三维点云数据修复算法-《计算机技术与发展》

文章信息/Info

Author(s):: LUO Lei; MA Rong-gui; MA Yuan; School of Information Engineering,Chang’an University,Xi’an 710064,China

Keywords:: point cloud processing; boundary extraction; hole repair; polynomial fitting; principal component analysis; radial basis function; clustering algorithm

摘要:: 目前车辆轮廓检测主要使用激光扫描技术,其过程为车辆低速通过龙门架,由安装在龙门架两端的测距雷达扫描车辆,在使用测距雷达获取车辆点云数据时,由于入射角过大,会在车窗处产生不正常的噪声点。针对车窗处点云数据异常的问题,提出车窗处点云数据修复算法。首先使用主成分分析法获得车辆点云数据的法矢,达到丰富点云数据信息的目的,之后将三维信息和法向量信息结合,利用聚类算法获得车辆侧面的三维数据和法向量,并且利用确定的三维数据和法向量的范围将侧面提取,结合曲面拟合去除特殊噪声点,再使用改进的 8 邻域深度差的方式提取车窗边界,最后利用径向基函数修复车窗点云数据。通过对多种车辆的车窗进行数据修复,验证了该算法的有效性。

Abstract:: At present,laser scanning technology is mainly used in vehicle contour detection. The process is that the vehicle passes through the gantry at a low speed and is scanned by the ranging radar installed at both ends of the gantry. When using the ranging radar to obtain the vehicle point cloud data,due to the excessive incident angle,abnormal noise spots will be produced at the window. Aiming at the problem of abnormal cloud data at the window,we propose a cloud data repair algorithm at the window.Firstly,the principal component analysis is used to obtain the normal vector of the vehicle point cloud data,so as to enrich the point cloud data information.Then,combined the three-dimensional information with the normal vector information,the three-dimensional data and the normal vector of the vehicle side are obtained by using the clustering algorithm,and the determined range of 3D data and normal vector will be extracted from the side,combined with surface fitting to remove special noise points. And then the improved 8 neighborhood depth difference will be used to extract the window boundary. Finally,the radial basis function will be used to repair the window point cloud data. We verify the effectiveness of the algorithm by repairing the data of various vehicle windows.