吴禄慎　俞　涛　陈华伟

(南昌大学机电工程学院　江西 南昌 330031)



基于自适应椭圆距离的点云分区精简算法

吴禄慎俞涛陈华伟

(南昌大学机电工程学院江西 南昌 330031)

摘要利用传统点云精简算法进行散乱点云简化会导致点云模型部分细节特征的丢失或模糊以及影响非平面区域的光顺性。针对这些问题，提出基于自适应椭圆距离的点云分区精简算法。首先，通过对邻域点集进行微切平面与局部曲面的拟合，计算出各点的法矢及曲率等；其次，利用所得几何特征信息，提取点云边界特征以及完成点云平面区域与非平面区域的划分；最后，采用改进后的精简算法对不同区域进行简化。实验结果表明，该算法不但能够快速完成符合要求精简率的数据简化，还能保护点云模型的细节特征以及保证模型非平面部分的光顺性。经过软件分析得出，精简后模型与原始模型的距离误差的标准偏差为0.015 mm。

关键词点云精简四元数最小距离法点云分区边界提取

REDUCTION ALGORITHM OF POINT CLOUD SEGMENTATION BASED ON ADAPTIVE ELLIPTICAL DISTANCE

Wu LushenYu TaoChen Huawei

(School of Mechanical and Electrical Engineering, Nanchang University, Nanchang 330031,Jiangxi, China)

AbstractApplying traditional point cloud reduction algorithm to reducing scattered point cloud will lead to missing or fuzzy of some detail features of the point cloud model and affecting the smoothness of non planar region. Aiming at these problems, we put forward the adaptive elliptical distance-based point cloud segmentation reduction algorithm. First, by fitting the tangent plane and local surface on neighbourhood set, it calculates the normal vector and curvature of each point; secondly, it uses the derived geometric feature information to extract point cloud boundary characteristics and to complete the partition of planar regions and non planar regions of point cloud; finally, it uses the improved reduction algorithm to simplify different regions. Experimental results show that the algorithm can not only rapidly accomplish data simplification in accord with the required reduction rate, but can also protect the detail characteristics of point cloud model and ensure the smoothness of non planar portion of model. Through software analysis, it is found that the standard deviation between the reduced model and the original model is 0.015 mm.

KeywordsPoint cloud reductionQuaternionMinimum distance methodSegmentation of point cloudBoundary extraction

0引言

逆向工程中，光学三维扫描设备被广泛用于精确、快速地获取物件表面坐标信息，但会生成大量的冗余数据，且点云数据一般散乱无序、十分密集。若将全部点云都用于曲面重构，不仅占用大量的计算机资源，操作效率低，而且庞大的数据不便于后续的存储、显示及传输等。因此，在保证精度的前提下对数据进行有效地精简显得尤为重要。目前，许多学者对点云精简做了大量研究，邵正伟[1]等先使用八叉树法对散乱点云进行点云空间划分，然后采用包围盒法[2]完成点云的均匀精简，能够快速完成精简，适合应用平面区域。……