Research On 3D Reconstruction System Based On Kinect And PCL

With the continuous development of intelligent robot, the robot's working environment is more complex,the requirements of the robot to awareness and understand the surrounding environment is more and more high. Robot vision has become one of the hot research all over the world. However,3D reconstruction technology is a key in the research of robotic vision, 3D reconstruction technology has practical application significance in the field of robot vision. On the reconstruction of an object or model of the surrounding environment by means of 3D reconstruction technology, then the robot can make judgment on the current environment and realize robot autonomous control or remote control. The means of obtaining point cloud data is expensive which has seriously hindered the development of 3D reconstruction technology application in various industries. Therefore this paper conducts the research in the technology of 3D reconstruction and put forward and builds a low cost efficient 3D reconstruction system base on Kinect and PCL. Each stage of this system use automatic programming processing, and put forward some improved algorithms of the reconstruction, then conduct experimental validation.This paper studied the current processing status of 3D data measurement technology and point cloud data processing, including point cloud acquisition, point cloud denoising, point cloud feature extraction, registration of point clouds, surface reconstruction, the main work is as follows:(1) Research on the current situation of the point cloud data acqui-sition,put forward a point cloud acquisition mean based on Kinect. Research the software system of Kinect, and programming to achieve access to the point cloud data which realize the low cost purpose.(2) Introduce PCL to this system first time and process point cloud by using PCI.The PCL is a study hotspot abroad, but domestic research is almost zero.PCL will be an important tool for3D reconstruction in future. The paper conduct the study of PCL developed the data acquisition program and data communication program based on PDA, and carried out the library.(3) Study the point cloud data denoising. Propose denoising method based on statistical distance for the difficulty in get rid of the nosie near and far away from the object. Realize effective removal of random point objects far and near. And then conduct effective reduction of the point cloud data base on the original structure.(4) Deeply study the theory of feature extraction from point clouds. Put forward the feature extraction method base on point cloud using point feature extraction, and realize it by PCL programming.(5) In point cloud data registration section:study ICP theory and achieve the function of point cloud data registration. In the surface rec-onstruction section:study the theory of the triangle subdivision pri-nciple. Use MLS to conduct point cloud data resampling and use greedy triangular subdivision to realize surface triangulation reconstruction. Implementation these process by programming. Finally obtain the 3D model with the help of some three-dimensional software.