Traditional forward engineering technology requires a lot of development experience.As the model is more and more complex, the development efficiency becomes lower and lower. Due to its short development cycle, high efficiency and low cost, reverse engineering has attracted more and more attention. At present, the reverse engineering technology has been gradually paid much attention in complex model, such as the automobile, aerospace,mold and process products. Although the application of reverse engineering technology becomes more and more widely, but there is still not a set of solutions which can perfect solves some important part of reverse engineering. This paper makes a deep research on the key technology of reverse engineering.In this paper, we face noise problems to combine the fitting ball method of APSS method and AABB tree method to complete denoising work. Through APSS method we can calculate spherical center and radius of the point clouds. Then we do ABBB-tree segment on the bounded sphere with curved surface so as to quickly find the bounding sphere which is the ray and the surface intersecting, then calculate the point of the ray and the surface and use these points as the point set after denoising. This method is proved to be highly accurate and efficient.Then we face the large point cloud problems to combine Bound Cocone method and the local feature size to complete simplify work. In Bound Cocone method, the interior and boundary points of point set are extracted by Cocone Bound. We calculate the negative pole in P by calculating the summit of Voronoi in k-field, through which we approach the middle axle in order to estimate the local feature size f(p) .Then we use Bound Cocone method to determine the interior as the central point, draw a sphere radius as ?f(p) and delete the rest points outside. Through the experimental comparison, the characteristics of the proposed method are verified in this paper.Finally, we face the outliers and noises of point cloud problems to combine RANSAC method and our simplify method to complete surface reconstruction work. This method uses the RANSAC to remove outliers .Then the method proposed in this paper to simplify the points cloud and fit spherical. At last by means of the Delaunay method, the surface is extracted and reconstructed from the triangular mesh. The experiment finally verifies that this method can effectively remove the outliers and noise and fulfill the surface reconstruction. |