Study On 3D Reconstruction Based On Point Cloud

As an important component of computer graphics, 3D reconstruction is a keyproblem in the field of reverse engineering, computer visualization, virtual reality,etc. With the development of science and technology, the traditional imagesbased 3D reconstruction algorithms can hardly satisfy the requirements ofmodeling and rendering of high precision and real 3D modules for their defaultsin precision, reconstruction speed and availability. However, point cloud based3D reconstruction algorithms can directly construct real 3D model throughscattered surface points. Currently, research on point cloud 3D reconstruction is afocus of 3D reconstruction, a difficulty in 3D reconstruction and a key to 3Dreconstruction.This paper got the meshes of objects using point cloud based 3Dreconstruction technology. The point clouds of objects were collected by 3D laserscanner. The surface meshes of objects were got from simplification andtriangulation of point clouds.To reduce the number of points for latter triangulation, we simplify scatteredpoint clouds. With the problem that the traditional K nearest searchingalgorithms'efficiency are not high enough to process vast space scattered pointclouds, a parameters controllable K nearest neighbors searching algorithm ispresented in this thesis. This algorithm searches points with step and deviation inthe middle of coordinate values of measuring points and uses the K nearest pointsin intersections. Finally, we use normal precision method to simplify point cloudsand keep more points in high curvature and little points in low ones.After getting the simplified scattered surface points of objects, we maketriangulation. With the problem that direct triangulation algorithms oriented tospace scattered point clouds are rare and inefficient and the qualities of targetmeshes are lower. This thesis proposes an improved mesh frontier algorithm todeal with space scattered point clouds. This algorithm sets up the criterion ofcandidate points and generates an initial triangle at first and gets 3D surfacemeshes of objects by repeating searching the best candidate point of extendableedge and judging whether the best candidate point is the best candidate point of the adjacent edges of extendable edge.Finally, we realize the scattered point clouds 3D reconstruction systembasing on the algorithms and criteria approved from this thesis and furtherdemonstrate the feasibility and validity of the presented algorithms.