Study On The Indoor 3D Map Generation Using Kinect

As an important component of computer graphics, 3D reconstruction is a key problem in the field of reverse engineering, computer visualization, virtual reality, etc. With the development of science and technology, the traditional two-dimensional map has inevitable defection the visual and vivid, which has been difficult to meet the requirements of the realistic 3D maps. And the 3D map generation technology based on point cloud can directly construct real 3D model through scattered surface points. Currently, research on point cloud 3D reconstruction is a focus of 3D reconstruction, a difficulty in 3D reconstruction and a key to 3D reconstructs on.Deep research on the registration and the reconstruction of the point cloud data is done.In this paper, encapsulation function in PCL and OPenNI module are used to collect point cloud data based on Kinect.Then the invalid point filter and the VoxelGrid filter is used to pretreatment. Then using normal distribution transform algorithm converts a three-dimensional point cloud data to continuously differentiable probability distribution function, and the point cloud data are registration. Finally with octree space integral mesh simplification algorithm and texture mapping generates 3d model based on point cloud data.In order to reduce the point cloud data for triangular mesh, this paper uses regional gravity compression method to do the sampling. After the point cloud data is contained in the outer region rectangle, the rectangle is divided into smaller rectangle. Then calculate the center of gravity and keep it and remove the redundant points, now point cloud simplification is completed. Due to the simplified three-dimensional point cloud data is a discrete distribution, hence the need for triangular mesh processing to obtain three-dimensional model.This paper presents an octree space integral mesh simplification algorithm to solve the problem of too many triangles. The basic idea of the algorithm is to determine the boundary voxels firstly, then insert them to the octree which is Branching on Need tree in this paper.Next to organize the data effectively the shrink operation is performed, and triangles drawn is used by the unit elements and super elements, which reduces the generations of triangles.Finally using the triangles forms the mesh, and completing the whole triangulation of scattered point cloud data.In this paper, VS2010 is used to do experiment. Point cloud data collection, point cloud filtering, point cloud registration and 3D model reconstruction based on point cloud are studied. 3d map of indoor scene are preliminarily realized.