3D Reconstruction, Restoration And Research And Algorithm Optimization

The repairing and optimization of3D model is an important question in the field of Computer Graphics. The reverse engineering technology is an important means to repair3D models. With the development of computer-aided geometric design, the reverse engineering technology is becoming the hot research spot at home and abroad. Reverse engineering includes data acquisition, data preprocessing, surface reconstruction. The repair of3D models in the reverse engineering’s applications has become a hot research topic in recent yeas. With development of laser scanner3D acquisitions and increasingly sophisticated hardware devices, we can get the point cloud data of complex models, however, due to the precision of the equipment, the objects themselves and other reasons, the models have a lot of noise, and the triangle meshes by surface reconstrution have some irregular triangles, and there are often data loss owing to partial covering of the part or limit of the measuring instrument within the acquired triangular models, so the models may be incomplete. Therefore, we need to smooth and repair the model in order to meet the subsequent application. In this paper, we focus on surface reconstruction, and the tranguler mesh smoothing and hole filling on mesh models.1. Based on the easy implementation of the regional growth reconstruction algorithm and the efficiency of search space data by the octree, a new mesh growing surface reconstruction algorithm based on octree is proposed. The algorithm can deal with point cloud data without the normal information and pre-progressing.2. Due to the laplace smoothing algorithm is easy to implement, but prone to lose features, we propose an adaptive and anisotropic smoothing algorithm for removing noise and keeping feature of triangular mesh model. First calculate the expectations normal of the model triangles, the normal meet an accurate representation of the desired direction as well as be similar to the original; then the offset and adaptive coefficient was obtained for each vertex; finally, we updated every vertex of the model. Experimental results show that the algorithm can avoid shrinkages very well and adaptive to preserve sharp features.3. A feature-preserving smoothing algorithm based on least squares is proposed. The algorithm proposes a quadratic energy function related to the optimized normal, and the parameters of the Gaussian function in the quadratic energy function are the average values of the angles between the triangle normal and the adjacent triangle normal, the Gaussian function is to distinguish the noise and the feature as a characteristic function, and then, inverse vertices and adjustments so that the vertices distribute more uniform. On the one hand, the algorithm make the model more smoothing except the feature regions, and the features are preserved.4. As the problem of hole-filling in the high-curvature region in the triangular mesh isn’t satisfactory, a new algorithm is proposed. In the mesh subdivision process, judge the mean curvature of the triangles that have a common edge, if the edge in the high curvature area, inserting new vertices in different ways according to the different situations of the edge length to achieve the mesh subdivision. Lastly, the filled mesh is refined according to surrounding mesh. The algorithm can make holes with neighboring mesh transit more natural and the repairing effect is satisfactory by many examples.