Camera Calibration And Three-dimensional Reconstruction Based On 3-parameter Model

Three-dimensional reconstruction is the important research contents of the computer vision and ultimate research objective. Three-dimensional reconstruction is mainly composed by these two steps: first, calibrate the camera, and then determine the camera extrinsic parameters. Once the camera has been calibrated, the camera extrinsic parameters between the images can be easily derived. Therefore, camera calibration is a key step in three-dimensional reconstruction. This thesis reviews the history of the camera calibration and the three-dimensional reconstruction, summarizes the mostly research production of these problems on both internal and overseas investigators. The main research contents and results of this thesis are as follows:1) An approach of camera calibration based on 3-parameter model has been proposed. The vanishing points constraint and theory have been used in this part, and then camera calibration can be obtained by the single image.2) A stratified reconstruction approach based on the uncalibrated camera is proposed to reconstruct a single cube. First, the fundamental matrix and infinite homography matrix is solved by the image information to projective reconstruction and affine reconstruction, and then the intrinsic parameters are solved by the three-parameter model. With these parameters stratified reconstruction can be carried out. At last, the object is reconstructed from the topological structure.3) A stratified reconstruction approach based on the uncalibrated camera is proposed to reconstruct Bezier curves and Bezier surface. For the Bezier curves reconstruction, reverse the control points of the curve and solve the fundamental matrix and infinite homography matrix by image information at first to projective reconstruction and affine reconstruction, then the 3-parameter model is used to solve the camera intrinsic parameters, so the control points of the curve can be stratified reconstructed. Lastly, a Bezier curve is used to fitting the control points. For the Bezier surface, the surface reconstruction is obtained by the Bezier curves reconstruction, and the approach is the same as the curve's reconstruction approach. This thesis's experiments are all based on the 3-parameter model of intrinsic parameters. There are some characteristics to use this model: It's simple to choose the 3-parameter model, the parameters are fewer than 4-parameter model or 5-parameter model. The demand of the target is not high with the experiment. And only need a single cube to calibrate camera and three-dimensional reconstruct, the solving process is linear. In this thesis, the OpenCV (Open Computer Vision) and OpenGL (Open Graphics Library) of VC++6.0 are used to experiment. And the experiment results show that it is feasible to use 3-parameter model to calibrate camera and three-dimensional reconstruct, and the algorithm has the better precision and certain stability.