An Improved ICP Algorithm for Point Cloud Registration

Abstract

In order to solve the problems that classical ICP algorithm has low registration efficiency in the process of point cloud data registration and high requirements for the initial position of point cloud, an improved point cloud registration method with ICP algorithm is proposed. A spatial registration scheme combining initial registration with accurate registration is adopted, using the FPFH algorithm to transform initial point cloud to an ideal position, which reduces the possibility of iterative nearest point algorithm falling into the optimal solution, Then, based on ICP algorithm, ISS algorithm is used to extract key points, and the search time of the nearest neighbor points is shortened by using the k-d tree structure, distance constraint and cosine constraint of normal vector angle are used to exclude wrong point pairs. By observing the experimental results we found that the running time and accuracy of the algorithm are obviously improved, which can meet the registration requirements and improve the registration efficiency.