Vision Based Mapping Of A Mobile Robot In Unknown Environment

The research of mobile-robot which involves much Knowledge such as computer vision, automatic control, Electronic technology and pattern recognition, is widely applied in space-exploration, military-service and room-service. In order to be more useful, mobile-robot is necessary to be autonomous navigation. Mobile-robot need to map the unknown environment for autonomous navigation. Though It is a mature solution to using laser or sonar sensors, visual-based SLAM by single camera is to be focus for low-lost and rich-information. In this work, we study the theory and technology of3D reconstruction in indoor environment, and implement an application.In this research there are some tough problems:extract feature from images, match feature from sequential images, renew camera’s pose in the3D-modle which caused by the mobile-robot’s moving.In this paper, we gave a detailed analysis of each step for the3D reconstruction. First, we introduce the basic theory of computer vision, camera imaging models and camera calibration methods. We calibrate the Kinect and give the results of the camera. Second, we extract, descript and track the features. And then we compare Harris, FAST and KLT algorithms, analyze the experimental data of the methods. Third, we depict the3D image with the3D points. At last, we estimate the camera position and orientation by the POSIT algorithm.Finally, we implements the slam-application based on mobile-robot developed by our lab. The result shows that the solution is effective and robust.