Theory And Technology Of Three-dimensional Reconstruction Based On Fault Data

3D reconstruction based on slice-images data is a main topic of visualization of 3D data sets. Reconstruction of 3D model is the premise of analysis and emulation. Based on this background, the paper mainly studies 3D reconstruction from cross-section data and related techniques. The method of tilling contours, and solving the key problems such as correspondence, branching, smoothness that directly affect the accuracy and applied value of 3D model are mainly included.While 3D models are reconstructed with surface rendering, triangles or polygons are needed to fill in the adjacent contours so as to form the surface of the object. With the optimized shape of triangles as objective, a method is presented to solve tilling problem based on Delaunay triangulation approach according to the criterion of the maximum of the smallest inner angles. While two smallest inner angles of available triangles are equal, the shortest path method which is adopted. The complexity of the presented method is low; it is easily implemented and reconstructed triangular grids are shapely. But the cone arises while the distance of the centers of upper cross-sectional contour and low cross-sectional contour are far from each other. In this paper, the above-mentioned problem is solved by aligning the center points of every contour. The shortcomings of the envelop box method such as great calculation and complex implementation are also overcome.Since every cross-sectional contours is likely to be nested and have different number, in order to ensure the correctness of geometry reconstruction, topology reconstruction should be carried out firstly before solid models are built, namely, cross-sectional contours are classified to determine their object. In this paper, nested trees that correspond with contours in every cross-section are built through the introduction of nested matrix. Then the minimum spanning tree of nodes with the same degree in the nested trees which present contours in two adjacent cross-sections is constructed. What is more, while constructing the minimum spanning tree, mutual overlapping area is used as constraining condition. The method facilitates the building of nested trees by the introduction of nested matrix, and then, the correspondences based on overlapping and that based on global are combined, which not only reduces the complexity of topology judgment but also determines the correspondence of contours accurately.As far as the same object is concerned, branching problem need to be solved while the numbers of contours in adjacent cross-sections are different. One of the solutions to the problem resulting from non-overlap multi-contours is to transform multiple branching problems into a group of single branching ones. Since the distance between the adjacent cross-sections is very close, upper contours and low contours bear some similarities. This paper puts forward a method that transforms multiple branching problems into some single branching ones according to the rate of perimeter of multiple contours. The method turns out to be feasible through experiments.Because of the sparsity of original data, a certain method to fit triangles orpolygons that will fill in adjacent contours is needed in order to reconstruct smooth surface. Finally, the triangular Bezier surface is reconstructed to fit triangular grid, which reconstructs shapely triangular grids and improves the smoothness of the surface.Experiments have demonstrated that the adoption of the above-mentioned methods to solve the problems of 3D reconstruction reduces the calculation and improves the accuracy of reconstruction as well.