基于点云的SUSAN特征点检测算法在三维重建中的应用

The Application of SUSAN Keypoint Detection Based on Point Cloud in 3D Reconstruction

针对三维重建中点云特征点检测问题,提出了一种基于点云的最小核值相似区(SUSAN)特征点检测算法,并将其应用于三维重建的初始配准.首先,对待测点云进行遍历,利用kd-tree数据结构获取三维r-邻域核值相似区,计算得到点云的候选特征点;其次,使用快速点特征直方图对候选点进行特征描述并实现两幅点云特征点间的匹配;最后,利用奇异值矩阵分解法计算变换矩阵,完成两幅点云的初始配准.实验结果表明该特征点检测算法计算效率较高,产生的特征点匹配准确,可为精确配准提供较好的初始位置.

Aiming at the problem of the keypoint detection algorithms in the process of 3 D reconstruction, a smallest univalue segment assimilating nucleus （SUSAN） algorithm based on point cloud is proposed and has been applied to initial registration in 3D reconstruction process. Firstly, the algorithm selected the candidate keypoints by obtaining 3D univalue segment assimilating nucleus of each point with kd-tree structure. Secondly, the features of key points are described by using fast point feature histogram （FPFH）. Then, we worked out the transformation matrix using singular value decomposition （SVD） method and got the result of initial alignment of two point clouds. Experiments show that the algorithm has high efficiency and can offer accurate matching of feature points and a good initial position for accurate registration.