Research On Registration Theory Of Three-dimensional Scattered Point Cloud And ICP Algorithm Of Gray Wolf

With the continuous progress of modern industrial technology and the rapid development of graphics application technology,3d reconstruction technology has attracted more and more attention in various fields.For example,in the industrial field,it can be used to detect whether the object parts meet the industrial standards;In the medical industry,it can be used to simulate human organs and find out what's wrong with patients.This technology mainly obtains the point cloud information on the surface of the object to be measured by scanning equipment and processes it in the computer,and finally establishes the three-dimensional digital model of the object.However,in the actual measurement process,due to the influence of the scope of the scanning equipment,the volume of the object,the measurement environment,etc.,only part of the data of the viewing angle can be obtained by one scan.Therefore,in order to obtain a complete object model,accurate registration of point cloud data measured from different perspectives is often required.As one of the important aspects of 3d reconstruction,the advantages and disadvantages of 3d point cloud registration algorithm will directly affect the final registration result.Especially now with the continuous improvement of production requirements,the emergence of massive point cloud data brings great challenges to the registration algorithm.At present,the existing registration algorithms have some shortcomings: on one hand,there are high requirements for the location between registration point clouds;on the other hand,with the rapid increase of point cloud data,the registration process takes a lot more time,thus the real-time performance of the registration algorithm is poor.In view of these problems,this paper mainly studies from the following point:First,introduces the common registration algorithm and in view of the shortcoming of existing registration algorithms in strict initial value of registration point cloud location,a rough registration method based on improved grey wolf algorithm was proposed before the precise registration.As a novel swarm intelligence algorithm,grey wolf algorithm has strong convergence and anti-interference.However,in order to make up for the tendency that the algorithm tends to fall into local optimality and accelerate the convergence of the algorithm,the nonlinear factor and difference operator are fused into the algorithm to further improve the performance of the grey wolf algorithm in the registration problem.Then,details the iterative closest point(ICP)principle and deficiencies of the algorithm.Aiming at the disadvantage of slow speed of searching corresponding point pairs in ICP algorithm,kd tree(k-dimension tree)was built to improve search speed.In order to make the ICP algorithm better deal with mass data processing,the uniform downsampling mechanism is introduced to reduce the number of alignment point clouds on the basis of keeping the features of the original objects.Finally,the double threshold mechanism is proposed to improve the ICP algorithm to avoid the influence of error points on the accuracy and efficiency of the algorithm.Finally,on the basis of the improved algorithm proposed in this paper,different types of point cloud registration experiments are designed.From the analysis of the experimental results,it can be concluded that the registration effect of the algorithm in this paper does not depend on the initial position between the point clouds,and it also speeds up the registration process.Besides,it has certain anti-interference to noise point pair,and has good adaptability and universality.