Reconstruction Algorithm Based On The Visible Shell Model Based On Two-dimensional Photographic Images Sequence Of Three-dimensional Polyhedron

With the development of multimedia and virtual reality, efficient 3D modeling algorithms whichcan realize the real-time computation are increasingly needed. 3-D model is essential to the object recognition, automatic design, video games, and digitalization of historical relics, reconstruction of historical buildings, and visual scene reconstruction.The traditional 3D modeling methods are based on the volumetric elements, and their complexities are proportional to the volume of object, which is not apt to the huge-volume objects. Since the result of a volumetric computation consists of disordered 3D points cloud, some triangulation methods which involve computing the distances between randomly scattered points are necessary to obtain the surface mesh of the object. How to avoid the redundant calculation and only to obtain the surface model that we are interested? The answer lies in the introduction of concept of Visual Hull, which is a hollow hull that contains the real object in world coordinate system. The unit of the surface model is a polygon instead of triangles; of course, the polygon may be just triangles. The normal direction of such polygons is the normal direction of the cone face, on which the polygons lie. When we get the surface model, we can project the vertex of polygons onto the input images to get the texture on the correspondent pixel and in turn get the texture on the surface model.The reconstruction in our system is based on silhouette information in photo sequence (Shape from Silhouette-SFS), and get polyhedral visual hull as the result of the intersection of cones correspondent to silhouettes. In this paper, I will present some algorithms of polyhedral visual hull reconstruction: (1 )Polyhedral v isual h ull generation a Igorithm b ased one dge-pool-searching, which constrains the 3-D intersections into the 2-D planes intersection via the epipolar theory to reduce the complexity a nd c an r educe t he t imes f or c ones t o b e p rejected o nto t he i mage p lanes, i n which we induced the index list to every points on the surface of object, and connected the related points to get sub polygons on surface and in turn get the mesh model, we replace triangulation process of disordered points cloud by edge-pool searching. ;(2)Reconstruction Based On Degenerated-Polygon-Intersection, which have the following improvements on traditional Bull Intersection: using degenerated polygon to organize intersected area on cone face, which may be discontinuous; a fast intersection algorithm of polygons on the cone face based on degenerated polygons, which is consistent between the continuous polygons and the discontinuous ones. The intersection of degenerated polygons is similar to the convergence of two ordered lists, which is comparably simple to the Bull operation; (3)reconstruction algorithm based on ray-tracing, in which all the rays that run through the vertex on silhouette of object will intersect with all the cones correspondent to other silhouette, and we have to calculate the intersection of all such line segments, the endpoints of which are considered to lie on the surface of object and we will get the point cloud on the surface of object. We can see all the calculation in ray tracing are on the rays, and the calculation fall into the intersection between face and line and intersection of line segments, which can be solvedby the ordinary geometry operation.All algorithms that introduced hi my paper focus on the hollow hull of object, and the calculations are a series of intersection of face and line that have no ply. And by employing the topology information that come with the points since their generation, we can get the surface polygonal mesh without triangulation and reduce the time used in reconstruction of visual hulls and help to realize real-time rendering.