车载激光扫描点云滤波方法研究

Research on Point Cloud Filtering Method for Vehicle Laser Scanning

针对传统的点云滤波算法存在阈值单一、地面点提取准确低的问题,本文提出了一种改进自适应阈值滤波算法。首先通过对点云数据进行二维投影并进行格网化处理;其次通过格网内最低点进行混合最小二乘曲面拟合;最后通过一级滤波阈值与自适应阈值实现非地面点滤波。为了对本文提出的自适应阈值滤波算法的有效性进行检验,分别使用城市中心道路与郊区道路点云数据进行算法实验。结果表明,本文提出滤波算法对城市中心道路点云滤波结果的一类误差、二类误差、总误差分别为4.6%、2.3%、3.7%;对郊区道路点云滤波结果的一类误差、二类误差、总误差分别为5.4%、7.1%、6.5%。相比于传统的移动窗口滤波算法,本文滤波算法无论是一类误差、二类误差还是总误差均更低,可准确区分出地面点与非地面点,表现出了更好的点云滤波性能。

Aiming at the problems of single threshold and low accuracy of ground point extraction in the traditional point cloud filtering algorithm,an improved adaptive threshold filtering algorithm is proposed in this paper.Firstly,the point cloud data is projected and grid processed;Secondly,mixed least squares surface fitting is carried out through the lowest point in the grid;Finally,non-ground point filtering is realized by first-order filtering threshold and adaptive threshold.In order to test the effectiveness of the adaptive threshold filtering algorithm proposed in this paper,the algorithm experiments are carried out using the point cloud data of urban central road and suburban road respectively.The results show that the first-class error,second-class error and total error of the filtering algorithm proposed in this paper are 4.6%,2.3%and 3.7%respectively;The first-class error,second-class error and total error of suburban road point cloud filtering results are 5.4%,7.1%and 6.5%respectively.Compared with the traditional moving window filtering algorithm,the filtering algorithm in this paper has lower class I error,class II error and total error,can accurately distinguish ground points from non-ground points,and shows better point cloud filtering performance.