安全多面体可视外壳及应用研究计算机科学与技术专业论文.docxVIP

Research Research and Application of Safe Polyhedral Visual Hulls Su Huanhuan(Computer Science and Technology) Directed by Associate Prof．Chert Guojun Abstract Recovering shapes of three dimension object from acquired real video images with geometric topology and texture information in complicate scene has long been all important research subject in computer vision，virtual reality and computer simulation．Several methods have been proposed to solve this problem，among which the method of Visual Hull(vH)or Shape-From-Silhouette(SFS)shows the best approximate shape of an object according to the silhouettes of images acquired with multi cameras in interactive rate，especially,the Polyhedral Visual Hull ffvI-I)produced a polygon surface model of visual hull，a view-independent approximation，with good quality and speed，and easy for acceleration by graphics hardware，therefore，was widely used in modeling and interactive systems． Nevertheless，when there were complicated objects with self-occlusion，or multiple objects with inter-occlusion in the scene，the reconstructed visual hull may exhibit phantom volumes， which could not be all accurate representation of objects in the scene．To solve the problem， after researching a large number of literature in the area of constructing the VH model and removing the phantom volumes，combining with the technology of the safe hull which Was a method only rendering the image of scene but not the intact model，we embarked on the following aspects of analysis and research in this paper： 1．A_f论r reseaehing the method of constructing the PⅦmodel，We订ansformed the three·dimensional spatial intersection operation into two-dimensional plane calculation，by introducing the geometric projection technology and epipolar geometry with incremental computation，then taken full advantage of optimized polygon clipping algorithm to reduce the number of intersection operation，finally constructed the view-independent three-dimensional model in real time． 2．On the