Research On Model Reconstruction Based On Single Scene Image

Image-based reconstruction has been the cutting-edge research problem of computer graphics. It embodies the theories and schemes of computer graphics, image processing as well as computer vision and acquires three dimensional modeling data from two dimensional clues of the image to realize the model reconstruction in the virtual scene. Thus, this technology has a promising future in the computer-aided design and reverse engineering.This thesis primarily works on the algorithms of image-based reconstruction and the realization of system from the perspective of single scene image. First of all, the thesis illustrates the acquisition of scene image and the methods of load and display for bitmap format file. Moreover, it establishes the schemes for computer vision application, which is on the basis of projective geometry, and proposes a new improved algorithm of vanishing point calculation from the idea of A. Criminisi vanishing point computation so as to build correct affine coordinates. Besides, the thesis studies the measurement of affine geometry on the image plane by the model of perspective camera and gives the mathematical methods for computing three dimensional coordinates of sparse points. These three dimensional points after computation can be shown in the virtual environment through the application of VRML and used for model reconstruction with the software of Pro/Engineer. Finally, the development of model reconstruction system based on single scene image is completed and the result of one specific modeling case indicates the correctness of algorithms and reasonability of system implementation.