Mobile Robotics Simultaneous Localization And Mapping Based On Stereo Vision

Simultaneous Localization And Mapping(SLAM)is a fundamental problem and hot area in mobile robotics,and the key issue is mobile robotics autonomous navigation in unknown environments.With more robotics were widely applied in extreme environment,robotics require higher positioning accuracy.Binocular vision sensor is closest to human perception,could provide abundant information of environment characteristics,SLAM based on binocular vision becomes hotspot in mobile navigation.This paper focuses the research on the mobile robot stereo vision SLAM system.The main work is shown as follows.First,the mobile robot stereo SLAM issues are explained in detail.The main work and structure of this paper is established.Secondly,the stereo vision SLAM and the two-wheeled mobile robot platform system in indoor environment are stated in detail.The platforms are classified as robotics motion model,robotics observation model and binocular camera model based on the requirement of mobile robotics stereo vision navigation system.By using Matlab,the stereo vision system was calibrated through Zhang calibration method.Thirdly,the extraction and matching algorithms of binocular vision system are studied.A new algorithm based on nonlinear diffusion filter in the stereo vision system,KAZE,was proposed.The introduction of nonlinear multi-scale space provides more accurate measurement information.KAZE and SIFT,SURF,ORB are simulated and compared,which proves the accuracy and stability of the KAZE.Finally,in order to enhance the accuracy of EKF-SLAM,a novel fuzzy neural network assisted EKF-based SLAM(FFN-EKF-SLAM)method is applied,which is combined fuzzy neural network and EKF-SLAM in stereo vision robotics.In the robotics simulation platform,the performance of FFN-EKF-SLAM and EKF-SLAM algorithm in different error scenarios are simulated,which proves that accuracy and robustness of FFN-EKF-SLAM is better than traditional EKF-SLAM algorithms.