Research On 3D Reconstruction Of Indoor Scenes Based On Kinect

3D reconstruction technology is an important research direction in computer vision and image and image processing.It has a wide application prospect in virtual reality,autonomous navigation of robot,industrial automation,reverse engineering and medical treatment.3D reconstruction technology is through the computer vision and computer graphics and other technical means of digital processing of the scene,and then reproduce the scene in the computer digital 3D model.In recent years,with the development of sensor technology and computer vision,the traditional 2D image no longer satisfies people’s pursuit of information diversification,so 3D reconstruction becomes a hot research problem.The emphasis of 3D reconstruction technology is on the acquisition of scene depth information and the processing of point cloud data.The low cost of depth information collection method and the high efficiency of point cloud data processing have become the focus of 3D reconstruction technology research.This paper mainly studies the method of 3D reconstruction of indoor scene using Microsoft Kinect.Kinect is inexpensive,suitable for a wide range of scenes,and can simultaneously access the depth of the scene information and color information,reducing the cost of 3D reconstruction of data acquisition.In the process of 3D reconstruction of indoor scenes,we focus on the existing algorithms applicable to the 3D reconstruction process,improve the shortcomings of the existing algorithms,from the data acquisition to the final 3D reconstruction of the scene to generate the entire process of the algorithm.In this paper,the working principle of Kinect,depth information acquisition mode,camera calibration principle and coordinate transformation relationship are introduced.The calibration of Kinect is completed by checkerboard algorithm.Secondly,we introduce the PCL point cloud library which is suitable for 3D data processing.The algorithm of the following point cloud data processing is based on the open source library and optimized for the shortcomings.Then the basic concept and storage method of point cloud and point cloud filtering algorithm are introduced.In the point cloud filtering algorithm,the two methods of filtering based on depth data and filtering based on point cloud data are introduced.Then,an important part of the 3D reconstruction process is introduced.Point cloud registration is to convert multiple points cloud data obtained from different viewpoints of the same scene to coordinate transformation into the same coordinate system,and register as a complete point cloud data.Finally,the experimental results show that the system can be used in Kinect.based 3D scene reconstruction,and the whole set of algorithms for Kinect.driven,data acquisition,point cloud data filtering and point cloud data registration are completed.Experimental results show that the proposed algorithm can achieve 3D reconstruction of indoor scenes,which has the characteristics of high operation efficiency and accurate reconstruction data.