Reverse Engineering Of Scattered Point Cloud Preprocessing Algorithms Research

Reverse engineering design technology which has developed in recent years is anew technology in engineering design. It has penetrated into products industry andbecome mainstream technology of modern product design. Reverse engineeringincludes three-dimensional data measurement, data preprocessing and modelreconstruction. Data preprocessing is the key step in reverse engineering design. Itsprocessing effect and efficiency directly affects the accuracy and efficiency of themodel reconstruction. This paper researches K-Nearest points search in datapreprocessing, boundary feature point extraction and data simplification of scatteredpoint cloud data. The research is very important to improve the efficiency and qualityof3D reconstruction.This paper analyzes and researches existing algorithms and improves existingproblems to provide more accurate feature information for three-dimensionalreconstruction. The main contents are as following.(1) Analysis two organization structure of scattered point cloud topology: octreeand K-D tree and gives an associated octree algorithm. This paper uses octreepartition method and the maximum number of data in a cube to decide whether toseparate the cube or not in data organization structure. Then, the uncertainty of lengthin octree partition method can be avoided.(2)Research two typical K-Nearest points search algorithm, discuss their searchaccuracy and completeness and analysis invalid search in search process. Then apruning hierarchical fast search algorithm on octree structure is proposed. Thisalgorithm can ensure the search accuracy and completeness through hierarchicalsearch and can improve the searching efficiency by pruning strategy. Experimentshows that the search efficiency of the algorithm is significantly improved ascompared with global search algorithm and hierarchical search algorithm.(3)Improve the problem of large amount of calculation and misclassification inexisting algorithm according to the advantages and disadvantages research of severaltypical boundary point extraction. This paper analyzes the distribution characteristicsof neighborhood points and designs two kinds of fast extraction algorithm forscattered-point-cloud boundary: the recognition algorithm based on pulling force andthe recognition algorithm based on reference plane. Experiment indicates that these two algorithms can extract the boundary of scattered point cloud quickly andeffectively. The extraction accuracy and efficiency is better than literature algorithm.Theoretical analysis shows that the algorithm can effectively simplify the calculationprocess and accelerate the extraction rate without affecting the extraction precision.(4) Analysis several simplification algorithms of scattered point cloud. In orderto solve the loss of geometrical feature points and boundary point in some algorithm,this paper designs a curvature simplification algorithm. Experiment indicates that thealgorithm can preserve geometric feature points of curve and remove more points on aflat area. The algorithm can solve threshold setting problem.