Research On The Binocular Vision SLAM For Mobile Robot In Indoor Environment

The simultaneous localization and mapping (SLAM) problem is a fundamental and hot problem in mobile robotics. It is one of the key problems for a mobile robot to be truly autonomous in an unknown environment. Vision sensor (camera) is the closest to the effect of human perception, and it can provide a wealth of environmental features and information. In recent years, with the advancement in image processing technology, the SLAM based on visual information is the current research focus in the field of mobile robot.In this paper, the binocular vision SLAM system of mobile robot in indoor environment are studied.Firstly, the SLAM problem of mobile robot are surveyed. On this basis, a systematic framework of the binocular vision SLAM is established, and it is also a research framework for this paper.Secondly, the extraction and matching process of SIFT feature is discussed. In this paper, through using the stereo matched SIFT features point as environmental visual landmarks. In order to reduce the complexity of matching in the SIFT algorithm, using 32-dimensional feature descriptors instead of the original 128-dimensional feature descriptors describes the SIFT image features.Thirdly, the linear calibration method based on a perspective transformation model is used to calibrate the binocular stereo vision system. The three-dimensional space information of the stereo matching point pairs is calculated by the three-dimensional reconstruction method based on linear calibration. In addition, the two-dimensional map and three-dimensional map of the indoor environment are discussed based on the point feature-map.Finally, the SLAM based on Extended Kalman Filter (EKF) algorithm is studied. The implementation procedure of the EKF—SLAM algorithm is presented. In order to solve the problem of the computational complexity in the EKF—SLAM algorithm, an approach of SLAM based on local maps is studied. For a mobile robot in indoor environment, the two-dimensional SLAM and three-dimensional SLAM simulation experiments based on point features are conducted in Matlab. The simulation results prove the integrity and realizability of the binocular vision SLAM framework.