散乱点云局部形貌标架量化及特征识别方法

Scattering Point Cloud Local Topography Frame Quantization and Feature Recognition Method

针对散乱点云特征识别结果存在噪声及特征遗漏的问题,提出一种基于曲面局部形貌标架的点云特征识别方法。基于点云局部中轴对样点的隔离作用,剔除样点欧氏邻域内的非测地邻域点,为曲面构造优化的局部样本模型。析取局部离散样本中的准共法截线点对集合,构造散乱点云的局部形貌标架。基于标架夹角的差异性,对曲面样本形貌进行量化分析,区分平滑、边界、棱边及尖角等特征区域,实现对中心样点属性的稳健判别。实验结果表明,该方法适用于不同采样密度的点云,可显著降低点云特征识别结果中的噪声点规模,且能有效减少特征遗漏现象。

A method for identifying the features of point cloud was proposed based on local topography frame of the surfaces to solve the problems of noise points and feature omissions in the recognition results of the existing methods.Based on the isolation effect of local central axis of the point cloud on sample points,the non-geodedic neighborhood points in Euclidean neighborhood of the sample points were eliminated,and the local sample model was optimized for the surface construction.The set of quasi-common normal section line point pairs in the local discrete samples was extracted to construct the local topography frame of scattering point cloud.Based on the differences of the included angles of the frame,the surface sample shape was quantitatively analyzed,and the feature areas such as smoothness,boundary,edge and sharp corner were distinguished,so as to realize the robust judgment of the attributes of the center sample point.Experimental results show that the method applies to point cloud with different sampling densities and may significantly reduce the scale of noise points in the point cloud feature identification results and feature omissions.