Reserch Of 3D Point Cloud Data Registration Based On Parallel Particle Swarm Optimization Algorithm

The application of 3D laser scanning technology in surveying and mapping is more and more extensive.More and more studies on it have been done.Its appearance broke the traditional measurement data acquisition and processing mode,which led to a change of the distance measurement technology.The research on the scanning data(point cloud)has become an important research direction.In this paper,we study the automatic registration of point cloud data directly from the scattered point cloud data obtained by measurements.Firstly,some basic knowledge related to the point cloud data is introduced,and the registration of point cloud data is discussed in detail.Then on the basis of summarizing and analyzing the domestic and foreign scholars' theories,this research combines the GPU parallel acceleration technique and the group optimization algorithm.In the study of global precision registration,a high precision,fast convergence and robust automatic registration scheme is proposed.Firstly,the research status and development trend of 3D laser scanning are studied,and the content,purpose and method of research are confirmed.Second,system introduced 3D laser point cloud data registration technology based,similar to other measurements,point cloud data related to the same coordinate frame,this paper analysis the transformation between coordinate systems in and point cloud data related to the coordinate system,and derives the relation.Third,based on the in-depth study of the existing group optimization registration algorithm and the classical ICP algorithm,after several experiments,summed up the advantages and disadvantages of each algorithm.Aiming at the problem that the point cloud registration algorithm based on swarm intelligence optimization computation time is long and the ICP algorithm is high to the initial position of the point cloud,a parallel particle swarm optimization algorithm based on CUDA is proposed.With two pieces of point cloud point to point distance the shortest for adaptation degree function,using particle swarm optimization algorithm,the particle is a natural parallel,will solve the adaptation degree is worth all thread operations assigned to the GPU calculated cloud point the relative position of the transform parameters.Then through coordinate transformation,the two point cloud unified into the same coordinate system.Because the GPU is not interfered with each other,it greatly improves the computing speed of the particle swarm algorithm,so that it can realize the fast and accurate registration of the point cloud.Fourth,the registration method based on,VS2010 and CUDA development environment is used to realize the 3D point cloud data of software registration,will be in different coordinate systems,mutual overlapping part of the point cloud data accurate registration to the same coordinate system.The experimental results show that this algorithm not only overcomes the shortcomings of ICP algorithm to the initial location of point cloud,but also effectively solves the problem of long computation time of point cloud registration algorithm based on swarm intelligence optimization.