基于改进随机抽样一致的点云分割算法

随着三维点云数据模型在三维建模、测绘、智能城市以及机器视觉等领域的应用,点云数据处理也成为一个研究热点。点云分割就是将三维空间中点云通过一系列算法,将散乱的点云数据划分成更为连贯的子集的过程,可以为后续的数据分析提供数据基础。针对随机抽样一致算法(random sample consensus, RANSAC)对杂乱、无规则点云数据分割效果不佳的问题,提出一种改进的RANSAC点云分割算法。该算法通过构建Kd(K-dimensional)树,利用半径空间密度重新定义初始点的选取方式,进行多次迭代来剔除无特征点,在实现点云分割的同时可以有效去除噪声点;此外,该算法重新设定判断准则,优化面片合并,可以实现点云的精确分割。实验通过对散乱点云数据进行分割,结果表明该改进RANSAC算法的点云特征提取数据量较大,面片分割的准确性较高,是一种有效的点云分割算法。

Point Cloud Segmentation Algorithm Based on Improved RANSAC

With the application of 3D point cloud data in 3D modeling, mapping, intelligent city and machine vision, point cloud data processing has become a research hotspot. Point cloud segmentation is the process of dividing the scattered point cloud data into more coherent subsets through a series of algorithms, which can provide the corresponding data base for the subsequent data analysis. To solve the problem that RANSAC algorithm is not effective in the segmentation of noisy and irregular point cloud data, an improved RANSAC point cloud segmentation algorithm is proposed. In this algorithm, KD tree is constructed, the selection method of initial point is redefined by using the spatial density of radius, the non-feature points are eliminated by multiple iterations, and the noise points are removed at the same time of point cloud segmentation; at the same time, the algorithm resets the judgment criteria, optimizes the combination of patches, and realizes the accurate segmentation of point cloud. The experimental results show that the improved RANSAC point cloud segmentation algorithm is a more effective point cloud segmentation algorithm, which has a larger amount of point cloud feature extraction data and a higher accuracy than Euclidean cluster segmentation algorithm.