Translated Abstract
With the development of laser scanning technology, we can acquire the 3D point cloud data quickly and accurately by using the laser scanner. The technology of 3D reconstruction based on point cloud plays an important role in more and more fields, such as historic preservation,medical imaging,digital city construction,so it has a highly research value.The key technologies of 3D surface reconstruction are deeply researched in this thesis,mainly including the pointcloud registration, point cloud simplification and triangular mesh generation algorithm of point cloud.
For cloud registration, firstly some traditional algorithms are introduced,such as classic iterative closest point (ICP) algorithm in the registration of two point clouds and the trim ICP algorithm based on the improved ICP algorithm.Then the principle and the operation efficiency of searching the corresponding nearest pointbased on k-d treeare analyzed. In order to improve the efficiency ofsearching corresponding points in registration, the traditional k-d tree searching algorithm is modified by adding a vector to save middle variable that can reduce the frequency comparison, so that the overall efficiency of trim ICP algorithm can be improved. For multi-view registration, firstly, a initial rough 3D point cloud model is constructed by using part of the overlap of information between adjacent point cloud . And then the multi-view registration is transformed into two point cloud resgistraion problem, by gradually modifying the positions of each point cloud in rough 3D point cloud model. Finally the more accurate point cloud model can be obtained.For solving the problem of point cloud simplification, firstly,the point cloud is divided into nonuniform bounding box by using the octree.Then data in each unit bounding box can be simplifyed according to the rule set in advance. Comparing with traditional bounding box method,our method can retain more information of the surface.Finally a better triangular mesh model has generated by using Crust grid division method.
At last, using MFC and OpenGL hybrid technology on VC platform, a point cloud 3D reconstruction visualization simulation system is designed and realized based on proposed approachs. After input scattered multi-view point cloud , the system can automatically reconstruct 3D surface mesh model.
Corresponding authors email
