结合形态描述和直方图特征改进的点云配准

Improved point cloud registration combining morphological description and histogram features

针对点云配准过程中配准效率和精度无法兼得的问题,提出一种将内部形态描述(ISS)特征点和二进制方向直方图(BSHOT)特征描述符相结合的改进的点云配准算法。该算法先采用体素格网下采样和ISS算法提取特征点;然后通过二进制的方向直方图(BSHOT)特征描述符结合汉明距离和随机采样一致性算法(RANSAC)进行粗配准;最终利用改进的ICP算法进行精确配准。利用多组点云数据对该方法进行验证。结果表明,在相同条件下,改进后的算法在配准时间和配准精度上均优于其他算法。说明所提出的方法具有较高的配准效率与精度,且随着点云数量的增多,配准精度的提高效果会增强。

In this paper, an improved point clouds registration algorithm combining intrinsic shape signatures(ISS) features and binary signatures of the histograms of orientations(BSHOT) is proposed to address the problem that registration efficiency and accuracy cannot be achieved simultaneously during the point clouds registration process. Firstly, the feature points are extracted by voxel grid sampling and ISS features;then, the BSHOT are combined with Hamming distance and RANSAC algorithm for rough registration;finally, exact registration is performed using the improved ICP algorithm. The results show that the improved algorithm is superior to other algorithms in terms of registration time and accuracy under the same conditions. It shows that the proposed method has higher registration efficiency and accuracy, and the registration accuracy is enhanced with the increase of the number of point clouds.