Three Dimensional Computer Vision for Computer Design and Manufacturing Applications M. Milanova, L. Nikolov, Svetlin Fotev* Department of Radiocommunication Technical University 83 Simeon Str. Sofia 1000 Bulgaria e-marl: mariofanna.milanova@infollnk.infotel.bg * Bulgarian Academy of Sciences Space Research Institute 6, Moskovska str. 1000, Sofia Bulgaria Aided Abstract Our work focuses on the problem of automatic generation of viewpoint invariant geometric surface models from range data to be used in CAD/CAM systems. We developed two methods for recovering the surface. First is on the base of Shape from Shading method and second is on the base of Facet method. A comparison between the results obtained using the 2 methods is made. We also offer new techniques for the representation and CNC processing of the surfaces. L Introduction Computer vision systems are gaining wide acceptance in industrial applications such as design automation and manufacturing automation. The problem of linking CAD and Vision systems is very difficult (Besl P,1985). We have to decide two important tasks: 1. How to gather reliable information about the 3-D object of the scene. 2. How to represent 3-D object for both CAD and Computer Vision applications. We propose surface based segmentation and parametric surface description of complex 3D objects liable to automatic mechanical process. The description of visible surfaces ob- tained by their recovering is used directly in the CAD~CAM systems where the offset of the object is determined easily and the tool path is simulated. 2. Methods for recovering of the objects Two methods are investigated for obtaining surface characteristics from intensity images in this work. 2.1 Modified Shape from Shading method First we use and modify Shape from Shading method and obtain range data information. We developed an algorithm for recovering three-dimensional shape of visible surfaces in a scene from a single two-dimensional image on the base of iterative method (Shirai Y.,1988). We restricted our attention to scenes composed of one opaque solid object bounded by piecewise smooth surfaces with no markings.