散乱点云分割技术研究与实现

点云分割是根据空间、几何和纹理等特征对点云进行划分，使得同一划分内的点云具有相似的特征。首先对获取的散乱点云数据进行去噪、填补空洞和畸变等预处理，然后计算最小包围立方体分割点云空间并构建八叉树加速邻域点的搜索，为每个点构造最小二乘邻域，分析散乱点云数据的高斯曲率和平均曲率，再通过区域生长法得到低噪声的精确分块，自适应、智能化地对点云进行分块。经实验验证，该方法可以获得较好的分割效果。

Research and Implementation of Scattered Point Cloud Segmentation Technology

This paper first carried on denoising, fill cavity and distortion of pretreatment for scattered point cloud data, and then calculated the minimum bounding box point cloud neighborhood spots and constructed the Octree search structure for each point least square neighborhood analysis of Gaussian curvature and mean curvature for scattered point cloud data. Finally, by the method of region growing, the low noise precision was obtained. The experimental results show that the method can obtain better segmentation results.