基于改进B-SHOT特征描述符的三维场景重建

3D scene reconstruction based on improved B-SHOT feature descriptor

针对三维场景重建中采用二进制方位直方图(B-SHOT)进行点云特征提取与匹配时,存在信息丢失导致匹配准确率降低的问题,文中采用一种改进的二值三维特征描述符DB-SHOT来进行特征提取与匹配以建立相邻点云之间的对应关系,结合随机抽样一致性算法(RANSAC)去除点云中的外点,并运用RANSAC的内点进行相邻位姿估计,进而进行相邻点云的融合。在保证匹配速度的前提下,解决B-SHOT信息丢失的问题,提高匹配准确率。通过采用标准数据集进行实验,验证了DB-SHOT作为三维场景重建特征描述符的可行性与有效性。

In allusion to the problem of low matching accuracy caused by information loss when the binary signatures of histograms of orientations(B-SHOT)is used in the 3D scene reconstruction to conduct point cloud feature extraction and matching,an improved binary three-dimensional feature descriptor DB-SHOT is adopted in this paper to conduct feature extraction and matching,so as to establish the corresponding relationships between adjacent point clouds. The random sampling consensus (RANSAC)algorithm is combined to remove the outliers of the point cloud. The interior points of the RANSAC are used to estimate the adjacent pose,so as to integrate the adjacent point clouds. The information loss problem of the B-SHOT is solved and the matching accuracy is improved in this paper on the premise of ensuring the matching speed. An experiment was carried out using the standard data set. The results verified the feasibility and effectiveness of taking DB-SHOT as the 3D scene reconstruction feature descriptor.