| Nowadays,artificial intelligence technology has gain rapid development,and market demand is increasing.The value of robot’s research and marketing becomes more visible.People hope that robot can complete different tasks more independently.Moreover,the information perception,localization and map building in unknown environment are the most basic abilities of the future robot.This paper designed a stereo VSLAM system for indoor robots.It uses the stereo camera to collect environmental information and realize localization in unknown environment by SLAM algorithm,and build the environment map.In this paper,modules of the system are theoretically studied,and experimental comparisons of different methods for each module have been made to make out a stereo VSLAM solution with good performance.At the same time,this paper also make experiments for the whole system in real environment,and the quantitative analysis results are given.Specific works are as follows:1)For stereo vision part,we make algorithm validation and experiments in real environment for binocular technology,and analyze Kinect device with experiments.Through technical index contrast and the analysis of the application scene,it shows that the stability of Kinect is higher,and achieve higher accuracy in current indoor application scene.2)SLAM part is based on sparse feature points algorithm,and technical evaluations have been made for each step including system calibration,feature point detection and matching,position calculation,and back-end optimization part.This paper chooses SURF feature points and using RANSAC algorithm for matching refinement.This system also build a loopback detection strategy and realize optimization with g2o library.3)For the whole system,a whole stereo VSLAM system for indoor robots has been built in this paper,and has made quantitative evaluations for the SLAM robot system based on Kinect.Result shows that this system’s error is less than 2cm,and the best application range is 0.8 meters to 6 meters,which can meet the demand of practical application. |