多因子激光雷达点云分区精简算法

激光雷达点云密度较大时会导致数据冗余，对点云数据的计算、存储及显示造成困难。本文针对激光雷达地形扫描点云的精简问题，提出了一种多因子分区点云精简方法。首先在改进点云组织方式的基础上，使用变异系数定权法并综合4种传统的点云特征提取因子，得到最终的综合评价因子，以划分特征点与非特征点；然后使用改进的八叉树将所有点依据其位置与数量划分为子集，并根据每个子集的特征点数量确定是否保留其中部分非特征点。该方法可更全面客观地对数据进行特征评估与选择，得到最具代表性的点，实现更高精度的精简。试验显示，多因子分区方法的误差比其他方法低20%～50%，且在整体试验区域精度的均匀性高5%～70%，证明该方法更优越。

Multi-factor LiDAR point cloud segmentation simplification algorithm

Excessive density of point clouds leads to data redundancy, making it difficult to calculate, store, and display point cloud data. Aiming at the problem of simplifying point clouds in LiDAR terrain scanning, this paper proposes a multi-factor segmentation point cloud simplification method. Firstly, on the basis of improving point cloud organization method, the variation coefficient weighting method is used to synthesize four traditional point cloud feature extraction factors to obtain the final comprehensive evaluation factor. Then feature points and non-feature points are distinguished by the final factor. Secondly, the improved octree is used to divide all the points into subsets based on their location and the point number, and determination on whether to retain some of the non-feature points are made according to the number of feature points in each subset. This method can evaluate and select the characteristics of the data more comprehensively and objectively to obtain the most representative points, and achieve more precise simplification. The experiment shows that the error of the results of the multi-factor segmentation method is 20% to 50% lower than that of other methods, and the accuracy is 5% to 70% more uniform in the overall test area, which proves that the proposed method is superior.