Kinect-based 3D Indoor Environment Reconstruction For Mobile Robot

The objective world has been digitized by 3D reconstruction so that the computer can operate and process. If the technology is being applied to the mobile robot, the robot can have the ability of perceiving the external environment. The ability can not only make the robot independent but also let it navigate and make a decision.The indoor environment 3D depth data collection is used by Kinect device in this paper and 3D reconstruction is completed from the multi-view of the indoor environment. First of all, it introduces the experimental platform of software and hardware and then introduces the Kinect data stream. Secondly, it introduces the depth image and point cloud data. Finally, the focus of this paper is the pair-wise registration under the multi-view. It makes full use of the registration parameters to realize stitching alignment and then complete the indoor environment 3D reconstruction.The pair-wise registration is realized by pattern matching of the extraction of feature point. The main work is as follows: 1) the environment point cloud data acquisition and preprocessing; 2) point cloud registration: based on KD-TREE data structure extract NARF key points and PFH features in the neighbor of NARF key points; match the PFH features according to the source and target data and compute the registration transformation matrix. 3) point cloud visualization; 4) experiments.In the experimental part three considerable experiments are given about point cloud registration and stitching and the scene is aligned by the reverse software, so it realizes the goal of 3D reconstruction. Experiments show that the method is fast and easy to implement with high rate of success.