Research On 3D Reconstruction And Motion Interaction Technology Based On Kinect

Because of the development of computer vision theory and 3D scanning technology in recent years,3D reconstruction has been a hot issue in the research at home and abroad. At present, due to the high complexity and inaccuracy of the stereo vision algorithm,3D reconstruction system with mature technology is mainly based on structured light scanning, but such systems are usually large and complex, expensive and difficult to operate. Once launched, with its low price, easy operation and other characteristics, Kinect depth camera greatly promotes the technique of 3D reconstruction to be plebification. On the other hand, released as the somatosensory device, Kinect also makes the way of motion interaction be possible. In this paper, based on the two functions of 3D scanning and motion interaction of Kinect, a 3D model reconstruction scheme based on Kinect is proposed firstly, then the interactive and display platform for 3D model is realized.The research work and innovation of this paper are as follows:(1) A depth image preprocessing algorithm based on improved bilateral filtering is proposed. Kinect is not a precision measuring instrument, combining with the depth and color information collected by Kinect, this paper can effectively filter the noise and fill the depth missing areas in depth maps.(2) A 3D reconstruction scheme based on improved KinectFusion is proposed. The previous work is mainly improved in two parts:1) In the process of point cloud matching, the ICP (Iterative Closest Point) iterative method used by KinectFusion can not estimate the motion components which are parallel to the scene surface in camera external parameter matrix, thus the result of this method is unsatisfied. The method which extracts edge feature points as auxiliary relations is dopted by this paper to improve the ICP iterative algorithm, which can effectively solve this problem.2) In the process of texture mapping, KinectFusion system does not have the corresponding description of the texture mapping operation. Through the color fusion volume method, this paper is able to paste texture for the model reconstructed, so as to get a more intuitive and realistic 3D model.(3) Combined with the skeleton tracking function of Kinect, the interactive display platform for 3D model was realized. This paper makes full use of the human joint points obtained by the skeleton tracking function of Kinect to recognize the posture, then we map the specific posture to the keyboard and mouse events, so as to rotate or zoom in the three-dimensional model rendered by OpenGLThrough the research work of this paper, ordinary users can easily and quickly use personal computer and Kinect camera to make a more accurate three-dimensional reconstruction for daily necessities or scenes. At the same time, with the realization of the interactive display platform for model based on Kinect, users can be able to operate on the platform and the model through the body posture. Therefore, the platform has a good interactive experience and can display items in multi-faceted, multi-angle or multi-level, which has practical significance in the cultural relics display, medical diagnosis, commodity display and other fields.