3D Reconstruction Based On Images And Meshing Algorithms

3D reconstruction based on multi images is always the hot topic.While3D reconstruction task is a problem which needs a lot of subjects tocombine the thinking,3D reconstruction needs much knowledge. In orderto acquire a more charming model, the problem of hole filling, simplify themodel and triangulation meshing are very important. This paper provides adeep research on the3D reconstruction and triangulation meshing.In this paper, some basic concepts related to3D reconstruction areintroduced, such as epipolar geometry, Homography Matrix, FundamentalMatrix and Essential Matrix. Then, several camera calibration algorithmsare presented. The bundle adjustment algorithm and its improvedalgorithm, which can reduce the iterated matrix size, are introduced indetail. This method can maintain the orthgonality of extra cameraparameter. Later, this paper proposes3D point cloud reconstruction, withthe improved Patch-based Multi-View Stereo algorithm, which canreconstruct more points and the point cloud model is accurate enough forthe point cloud calibration step. At the same time, new designed cloudcalibration algorithm is also introduced，which is convenient for watchingand used for reprojection back to image to form range image. After that,this paper introduces the meshing algorithm, mainly focuses on the pointcloud we construct, and after full consideration of the new generated pointcloud’s feature, we design an improved Poisson surface reconstructionmeshing algorithm. At last, this paper presents two texture mappingmethods used in the research, direct choosing algorithm and maximizationarea projection algorithm.