Research On The Correlation Problem Of SLAM Based On Kinect Monocular Vision

Simultaneous Localization and Mapping is the most important development research direction in the field of robot that can move independently in recent decades.SLAM technology is the core technology of autonomous robots,used in robot navigation,control,testing,production and other aspects.Especially in the 21 st century,the visual sensor that can acquire the unknown environmental information more accurately and accurately becomes the core technology of SLAM.Some breakthroughs have been made in the theory and method,and it is gradually moving from laboratory research to mature market application.Autonomous mobile robots based on visual SLAM require high real-time and robustness in real-life and production,.So for the SLAM of visual,real-time and robustness become the latest research hot spots.In this paper,we use the Kinect for Windows of Microsoft Corporation as environment-aware sensor to study the mobile robot's SLAM.Based on a large amount of excellent literature both at home and abroad,this paper summarizes several classical feature extraction and feature matching algorithms,and proposes a more real-time and robust M-FAST feature point extraction and matching algorithm Apply this algorithm to RGB-D SLAM.The main contents of this article are as follows:(1)The basic principle and general process of SLAM algorithm are briefly introduced.The current research status of SLAM technology at home and abroad is introduced and researched.The hot spot SLAM technology of SLAM is researched deeply.In view of the current SLAM research,the visual method of monocular SLAM based on Extended Kalman Filter(EKF)is mainly discussed.(2)The research on how to get the visual signposts in SLAM algorithm system is studied.For common feature point detection algorithms,such as Harris feature point detection algorithm,SIFT feature point detection algorithm,SURF feature point Detection algorithm,FAST feature point detection algorithm for analysis and comparison.According to the real-time feature of 3D reconstruction in SLAM algorithm,the fast feature point detection algorithm(FAST)is selected to obtain the natural visual landmark in the unknown environment to correct the accumulated error of SLAM algorithm in real time.FAST algorithm is optimized,and a M-FAST algorithm is proposed to improve the accuracy,real-time and robustness of corner detection.SURF algorithm is used to match feature points and improve the robustness of feature matching.(3)In this paper,the monocular vision SLAM algorithm based on extended Kalman filter(EKF)is studied in depth,including the principle of EKF,the establishment of SLAM model,the construction of SLAM framework based on EKF monocular vision,and experimental analysis And verification.And introduce Kinect depth camera to obtain the two-dimensional information and depth information of unknown environment,detailing the basic working principle of Kinect depth camera.(4)This article describes the latest two years,the study of visual SLAM algorithm,based on Kinect RGB-D SLAM algorithm and ORB SLAM algorithm.The M-FAST corner detection and SURF feature matching algorithm was applied to the Kinect-based RGB-D SLAM algorithm.