基于多视角区域生长的复杂点云模型分割

为提高三维点云模型在特征模糊区域的分割精度,提出了一种借助多视角区域生长的分割方法。基于网格法向量方向相异性原则,初次将模型划分为不同类别的子区域,在相应区域建立点云与多视角距离图像的一一映射关系。利用Canny算子对灰度的敏锐性获取独立连通域并计算其重心坐标,根据对应关系在三维点云中提取对应点作为种子点,然后引入网格法向量的偏移角度分离邻接面,同时对剩余彼此独立的分割面按照迭代搜索最近点的原则进行提取,并利用KNN算法去除离群点实现分割优化。在选取的模型数据集上进行实验,结果表明该方法能够实现复杂点云模型的合理划分,分割精度不低于80%。

Complicated Point Cloud Model Segmentation Based on Multi-view Region Growing

In order to improve the segmentation accuracy of the 3D point cloud model in the feature fuzzy region, a segmentation method based on multi-view region growing was proposed.Based on the principle of direction difference of normal vectors of grids, the model was divided into different categories of sub-regions.Then the one-to-one mapping relationships between point cloud and multi-view distance images were established in the corresponding regions.The sensitivity of Canny operator for gray level was used to obtain independent connected domains and their barycentric coordinates were calculated. The corresponding points were extracted as seed points in 3D point cloud.To separate the adjacent surfaces, the offset angle of normal vectors of grids was introduced.At the same time, the remaining independent surfaces were extracted according to the principle of iterative nearest points.To achieve segmentation optimization, KNN algorithm was used to remove the off-group points.Experiments were carried out on the selected model data set.The results showed that the complicated point cloud model could be divided reasonably by the proposed method, and the segmentation accuracy was not less than 80%.