针对传统特征点提取算法存在的对噪声敏感、人机交互效率不高等问题，提出一种基于点签名的散乱点云特征点检测方法。借助于k邻近方法，统计所有散乱点云的点签名，并给出相应的点签名直方图与色阶图。基于正态分布的规律设定曲率阈值，完成特征点云的初步筛选。在初步筛选的基础上，对候选特征点云在其主曲率方向上进行投影，完成特征点的进一步筛选。实验结果表明，该特征点检测方法适用于各种不同的点云数据，能够准确、有效地提取出特征点，且人机交互效率与抗噪性能均优于传统k邻近方法。%The traditional method for feature point detection from 3D scattered point cloud is sensitive to the noise data and has a low efficiency of human-computer interaction. To deal with the problems, a method based on point signature is proposed. The point signature information of all the points is calculated depending on the k nearest neighbors, the related histogram and color-level chart are drawn. A threshold is built according to the Normal distribution to finish the preliminary screening. The candidate feature points are projected in the directions of principle curvatures to select the feature points again. Experimental results show that the proposed method is suitable for various kinds of points models, which can detect the feature points accurately and effectively, and the efficiency of human-computer interaction and noise immunity are better than the traditional methods.
展开▼
