Research On Visual SLAM Algorithm Based On ROS For Indoor Scenes

In recent years,with the rapid development of artificial intelligence technology,mobile robots have been widely used in various fields of social life.Autonomous mobility is an important development requirement for the application and promotion of mobile robots in various industries.Simultaneous Localization And Mapping(SLAM)is considered to be the key to achieving autonomous mobility of robots in unknown environments.This paper mainly studies the visual SLAM algorithm of mobile robots in an indoor environment.Aiming at the problems of low efficiency and large error in the traditional visual SLAM method,this paper studies the front-end and back-end separately,and makes improvements to obtain an indoor vision SLAM method with good accuracy and real-time performance.First,after studying the basic knowledge of the Robot Operating Platform(ROS)and the overall framework of visual SLAM,the construction of the ROS-based visual SLAM experimental platform was completed,and then the imaging model and camera calibration method of Kinect camera were studied,and In the ROS environment,Open CV is used to calibrate the internal and external parameters of the color camera and depth camera to align the RGB images and depth images collected by the camera.Secondly,the three links of the front-end of visual SLAM: feature extraction,feature matching,and motion estimation are studied.Aiming at the problem that the traditional ICP algorithm cannot accurately solve the pose of the camera due to the lack of pixel depth data in the motion estimation process,a mixture of iterative closest point(ICP)and efficient n-point perspective(EPn P)algorithms is proposed for motion estimation,which further improves the accuracy of pose estimation.Then,in the back-end of visual SLAM,in view of the shortcomings of traditional random loop detection algorithms,such as high complexity and time-consuming,a loop detection algorithm based on visual bag-of-words model is introduced to quickly and accurately reduce the cumulative error during positioning,then the graph optimization model is studied,and the global graph optimization(g2o)is used to optimize the global pose,and the globally optimal camera pose and motion trajectory are obtained.Finally,the evaluation tool provided by the public data set TUM is used to evaluate the indoor visual SLAM algorithm before and after improvement,and it is verified that the visual SLAM method proposed in this paper improves the accuracy of the system while ensuring the real-time performance of the system.In addition,experiments were performed in a real scene using the ROS-based visual SLAM experimental platform,and clear point cloud maps and octree maps and estimated motion trajectories are given,which verifies the effectiveness of the improved method.