Nowadays, the methods used to gain point cloud are sensitive to the surface feature of measured ob -jects.As a result, plenty of holes and invalid points are inevitable in point clouds .An efficient point cloud re-construction can be significant for the point cloud analysis and application .This paper studies the filtration of the training points used for point cloud reconstruction .By choosing accurate points for training , less error points af-fects the reconstruction , and improves the training efficiency greatly .The improved K -nearest neighbor algo-rithm , which combines with the object ' s structure feature , is introduced into reconstruction .These two improve-ments can help with a much more actual point cloud reconstruction .%现阶段三维点云的数据采集方法因其对物体表面的敏感性,其测量结果中不可避免地存在大量的空洞、毛刺区域.三维点云的修补对于点云数据的进一步分析与应用具有十分重要的意义.针对三维点云数据的修补问题,对训练数据进行分类筛选,避免了误差点对于点云模型修复的影响;结合点云数据的拓扑结构特征,将改进的K最近邻法应用于点云模型修复,得到了贴近实际的点云模型.
展开▼