Application Of Kd-Tree Algorithm Based On BBF In Point Cloud Registration

With the development of reverse engineering,three-dimensional point cloud registration technology as an important branch of reverse engineering technology has also been more and more widely used.The purpose of 3D point cloud registration is to model and obtain a complete target object.However,in the actual measurement process,due to the existence of objective factors,only partial view data can be obtained.To obtain a complete object model,different perspectives must be measured.Get the point cloud data for registration.The advantages and disadvantages of the registration algorithm will directly affect the final registration results.So more and more researchers have improved the registration algorithm in different directions.Especially with the current requirements of production requirements,the emergence of massive point cloud data brings the algorithm efficiency and accuracy requirements to the registration algorithm.ICP algorithm has been widely used in point cloud registration because of its excellent registration principle.However,the current ICP algorithm based on kd-tree consumes a lot of unnecessary information in the registration process because of the unrelated search problem of kd-tree.time.In view of the deficiencies of this point,this article has started to work on the search-independent problem of the optimization algorithm.Aiming at the problem that kd-tree search algorithm has low efficiency of backtracking query,this paper proposes an improved algorithm based on the principle of BBF algorithm.It can change the backtracking structure of the search algorithm by establishing a priority sequence to store the query path points,thereby reducing the algorithm backtracking time and improving the algorithm.effectiveness.According to the existing research,the retrospective search time of the kd-tree search algorithm accounts for 80% of the entire search time.Through experiments,it is proved that by changing the search algorithm optimized by the backtracking structure,the search efficiency of the corresponding point pair can be effectively improved,and the research has certain research.value.In the face of the problem of low registration of massive point cloud data,the ICP iterative closest point algorithm based on KBF improved kd-tree combined point cloud registration was used for registration.The improved algorithm was compiled on the experimental platform,and the improved algorithm was compiled into a subclass function according to the PCL subclass compilation rules.The function was called on the ICP accurately and on time for the registration of massive point cloud data.Experiments show that the improved kd-tree algorithm improves the iterative efficiency of ICP,reduces the time for registration,and proves the effectiveness of the improved algorithm in the registration of massive point cloud data...