Three Dimentional Registration Of Point Clouds Based On Total Least Squares And Improved ICP Algorithm

As an alternative approach for data acquiringV3 D laser scanner can obtain 3D point cloud data of objects or scenes quickly and efficiently. Focusing on point cloud registration, the thesis studied three-dimensional datum transformation and improved ICP algorithm based on total least squares5TLS6 and robust estimation.1. A three dimensional model with constraints and robust algorithm was designed and studied for three dimensional datum transformation of point clouds. Within consideration of leverage and its effects on weight function and total least squares model, formula derivation was provided. Three robust functions, such as Huber, IGG and IGG III were put forth for 3D point cloud coordinate conversion. Experimental results indicate that the model based on the total least squares has better performs than the least square, moreover, both IGG and IGG functions are more robust than Huber.2. The three dimensional datum transformation model based on multivariate total least squares was researched. For dealing with model errors and ill-posed problem of three-dimensional datum transformation model with constraints, the transformation model based on the multivariate total least squares was proposed within consideration of the symmetry after being processed by the center of gravity. Its solution was also provided. The experiments show that the model based on multiple total least squares has better precision and efficiency than the traditional solution.3. Improved ICP algorithm with the multivariate total least squares was adopted to solve the parameters of rigid body transformation. The feasibility and effectiveness of the improved algorithm were analyzed. After obtaining initial position of point clouds with PCA, searching the closest points by KD-Tree, the improved ICP algorithm based on the multiple total least square algorithm was adopted to solve the rigid body transformation parameters. Comparing with the ICP algorithm based on the four element method, the experiments demonstrate that the improved ICP algorithm has better efficiency.