Research And Design Of Multi-view 3D Reconstruction System Based On Stereo Vision

Three-dimensional reconstruction is one of the classic problems in computer vision research.in this paper,based on the theory of stereo vision,while introducing the structure from motion(SFM)algorithm and multi-view depth map fusion algorithm to achieve a three-dimensional reconstruction of multi-view system based on stereo vision.Three-dimensional reconstruction system consists of three modules:Depth map acquisition module based on stereo vision depth map,Motion recovery module based on the SFM and a multi-view depth map integration module.Depth map acquisition module based on stereo vision Zhang Zhengyou calibration method is adopted to calibrate the camera.The Bouguet algorithm which relies on External camera parameters is used to fulfill the stereo rectification.In the Stereo matching area an improved local stereo matching algorithm based on B-T matching cost is proposed.And by analyzing the experimental results,the detailed comparison results show that,the improved stereo matching algorithm is improved in accuracy.Motion recovery module based on SFM Using the SFM algorithm based on optimization.For the three-dimensional reconstruction system,some optimization have been made.Experimental results show that the SFM algorithm based on optimization recovery more accurate motion matrix,meet the needs of multi-view depth map integration module.Multi-view depth map integration module Using a depth map fusion module based on the visibility constraint,the module has a series of consistency detection based on the visibility constraints on the depth map with a lot of mistakes and confliction,and finally,integrate the multiple views into one robust depth map reconstruct a full view 3D point cloud model consistently.The final reconstruction experiment got a high accuracy result,show that the role of the depth fusion algorithm.The innovation of this paper is mainly on two aspects,one is to propose an improved matching algorithm B-T based on local stereo matching costs,the experimental results show that this algorithm is superior to other local algorithm in accuracy.The other is the improvement on the original SFM algorithm based on optimization engine,the robustness and the efficiency of the algorithm are improved.In addition,the 3D reconstruction system adopts multi view depth map fusion algorithm,which guarantees the robustness of the algorithm for the three-dimensional reconstruction,and makes the reconstruction effect more perfect.