Affiliations: [a] Faculty of Mechanical Engineering & Automation, Zhejiang Sci-Tech University, Hangzhou, China
| [b] College of Mechanical and Electronic Engineering, Northeast Forestry University, Harbin, China
Abstract: The traditional point cloud alignment method suffers from drawbacks such as a large extensive computation, low computing speed, and poor alignment accuracy. To overcome these problems, this paper proposes a fast and highly accurate algorithm based on fast point feature histograms (FPFH) algorithm and spatial constraints. The proposed algorithm first filters, denoises the point cloud dataset, and calculates the point cloud normal to obtain the FPFH eigenvalue. Then, the vertebral space is divided into three regions according to its location, and the feature points in each region are calculated. The Euclidean distance between a feature point and the boundary of the adjacent region, and the weight coefficient corresponding to the feature point are given according to the calculated distance. The method overcomes the defect that the query feature point has a large workload in the traditional ICP algorithm and improves the registration precision of the point cloud. The experimental results show that the proposed method effectively solves the problems of the traditional point cloud registration algorithm, can effectively reduce the mismatch rate of point cloud registration, and can improve the registration accuracy and stability without reducing the registration of the elements.
