Indoor Positioning And Mapping Technology Of Mobile Robot Based On RGB-D Camera Research

Simultaneous localization and mapping(SLAM)technology is an important technology to solve the autonomous motion of mobile robots.When the robot is in an unknown indoor environment,it can no longer obtain its own position information through external.For this problem,RGB-The D depth camera is a good solution as a visual sensor.Based on the classic SLAM framework,this paper analyzes the composition of the Kinect1 depth camera,the camera imaging model,the working principle of depth measurement,and completes the calibration of the camera.The features of the image were extracted by SIFT,SURF and improved ORB feature extraction methods.The extraction speed of the three methods was improved.The improved ORB feature extraction method was better than the former two.The improved ORB feature extraction method was used to optimize with FLANN.Matching method to achieve fast extraction and accurate matching of image features.The pose of the camera is estimated based on the matched feature points.When the 3D position of the matching point is known,the PnP method can be used to solve the camera pose.The multi-point perspective PnP method for solving poses has linear and nonlinear methods.In this paper,the linear P3 P method is used to solve the camera pose matrix,and the nonlinear pose analysis is used to solve the camera pose matrix.In the method of estimating the pose of two images,the difference of the pose matrix obtained is not large,and finally the initial positioning function of the camera can be realized.However,the P3 P method solves the problem on a large scale by using only three pairs of matching points.The error in the pose problem is relatively large,so it is better to solve the pose by the nonlinear Bundle Adjustment method in the SLAM process of the system.Analyze the method of back-end optimization,and the map effect solved by the pose map optimization is better than the unoptimized processing.The word bag model is used for loop detection and similarity calculation to solve the problem of whether the mobile robot passes the same place in the SLAM process.Finally,the map construction problem in the SLAM process is solved.The set of estimated camera position points is used as the sparse point map.On the basis of obtaining a sparse map point,a three-dimensional frame cloud is constructed and the point cloud map is stitched to construct a three-dimensional image.Dense maps,realizing mobile robots in indoor sparse point map construction and dense map construction.Finally,build a software and hardware platform,carry out simple three-dimensional modeling of the Turtlebot2 mobile car,and introduce some components to establish the equation of motion of the car.The algorithm was tested on the Turtlebot2 mobile car,eventually achieving the goal of positioning and building dense three-dimensional maps for mobile robots.