针对现有点云多要素分割算法分割面状点集时分割精度低、分割块合并效果差等问题,提出了一种改进的多规则区域生长算法。一方面,计算点云数据的平面拟合残差,基于平面拟合残差设置种子点条件,对面状点集分割进行优化,以此提升面状要素分割的精度;另一方面,在距离条件的基础上,结合相似性和体积变化条件对合并策略进行改进,以实现分割块的有效合并;此外,利用中位数、Baarda数据探测法和k均值聚类分别对算法中涉及的阈值参数进行自适应设置。采用三种不同类型的点云数据进行实验,结果表明:改进算法能够提升面状点集的分割精度,提高了分割块合并的准确性;与其他算法相比,改进算法能够同时兼顾精度和效率,分割结果更具优势。

With regard to the low segmentation accuracy of planar point sets and poor merging effect of segments in the existing multi-factor segmentation algorithms of point clouds, an improved multi-rule region growing algorithm was proposed in this paper. On one hand, the plane fitting residuals of point clouds were calculated, based on which, the seed condition was set and the segmentation of planar point sets was optimized, so as to increase the segmentation accuracy of planar factors. On the other hand, on the basis of the distance condition, the merging strategy was improved in combination with similarity and volume changes to achieve effective merging of segments. In addition, the threshold parameters involved in this algorithm were set adaptively using the median clustering, Baarda data snooping, and k-means clustering. Furthermore, three different types of point clouds were tested, and the results show that the improved algorithm can boost the segmentation accuracy of planar point sets, and enhance the veracity of segments merging. Compared with other algorithms, the proposed algorithm can take into account both accuracy and efficiency and has better segmentation results.