Researches On Several Problems Of 3D Reconstruction Of The Computer Vision

The computer vision is an important branch of computer science and artificial intelligence. With the development of the science and technology, the application of computer vision has been appreciated extensively. One of the main componets of the computer vision is the reconstruction of 3D world from the 2D images. Most of the 2D images we see have been derived from this 3D world by camera systems that employ a perspective projection to reduce the dimensionality from three to two. Although the spatial information of the 3D object is perhaps containedined in the 2D image, in order to understand the nature from images, the process of getting 3D information from 2D images is necessary.The main study contents of computer vision include input device, low level -vision, middle level vision, high level vision, and system architecture. Under the digital camera Condition,this dissertation maily studies the issuses of edge detection,features detection and primitives of reonstruction about the computer vision,and proposes some novel and effective methods to cope with those problems.Edge detection is the basic research of 3D reconstruction of computer vision,and also is one of the most important research direction of the computer vision,edge detection may have a direct influence on 3D reconstruction effect. This paper put forward a novel image edge detecion methord based on uniform B-spline with shape parameter by modifying profit and loss data.a smoothing surface of the digital image is present by the new modified data.the edge point is detected by either computing the local extreme of the directional derivative or computing zero-crossing of the second order directional derivative of the smoothing surface.Experiments demonstrate that this algorithm is simple,accurate and steady.It can wipe off the bogus edge commendably and process the digital image in real time.Corner feature is the important feature of the image.It plays an important role in the field of computer vision. A method to detect corner points of digital curves is presented.Fistly,B-spline with shape parameter model is used for approximating the original curve so that the curvatures can be estimated with the analytical expression and then the corner points can be detected by the curvature distribution. Moreover the B-spline with shape paramete approximation improves the precision of the curve fitting and also assures the veracity of the he curvatures. So the precision of the corner detection is insured. Experimental results show the validity of this method.Points,line segment are mainly used as the geometric primitives of the stereo matching in the traditional 3D reconstruction. This geometric primitives can't effectively represent the free-fromed curves, so accurate reconstruction of free-formed objects was very difficult. A algorithm based on CB-spline curves is presented for 3D reconstruction of 3D objects. CB-spline Curves which have the property of the affine invariance can be used as primitives to reconstruct 3D objects under the assumption of the affine camera model. Experiments demonstrate that this algorithm presented is more accurate and robuster than other approaches.A new model of k order exponential uniform B-spline with shape parameter is given. Most of its properties similar to those of the B-spline model, and also it has a shape parameter. The spline curves which are constructed by the exponential uniform B-spline with shape parameter can adjust the curve's shape and the degrees between the curve and its control polygon with changing the parameter value. The new model can be applied to the field of CAD/CAM and computer vision.