Monocular Visual Odometry Fusing ArUco Markers And ORB Feature Points

Visual localization and navigation will be an important navigation mode for mobile robots to achieve full autonomous movement in the future.But at present,the research on machine vision is far from meeting the needs of practical application.For this reason,a visual localization method based on ArUco markers is proposed in this thesis.Aiming at the shortcomings of current visual SLAM,a monocular visual odometry fusing ArUco markers and ORB feature points is designed,which can make the mobile robot obtain its absolute posture in the fixed world coordinate system to achieve navigation.The specific research contents of this thesis are as follows:The principle of estimating camera pose by using a single ArUco marker is analyzed,and the accuracy of obtaining camera pose based on a single ArUco marker is verified by experiments.In order to meet the localization requirement of mobile robots,multiple ArUco markers randomly distributed in space are reconstructed.Then,the unique reference world coordinate system is determined,and the 3D coordinates of all the corner points of ArUco markers in the world coordinate system are obtained.Experiments show that the camera’s absolute pose in the world coordinate system can obtained based on the reconstructed markers’map when any ArUco marker seen in the space,which provides an important method for the localization of mobile robots.Considering the limitation that the camera’s pose can not been estimated when the ArUco markers is not visible,the monocular visual odometry based on ORB feature points is deeply studied in this thesis.The origin of scale uncertainty and scale drift in monocular ORB-SLAM is found by analyzing the initialization process of monocular VO and the principle of camera pose estimation.The analysis shows that the combination of ArUco markers and mono VO can not only overcome the limitations of localization based on ArUco markers,but also solve the scale’s drift and uncertainty of mono VO.Therefore,the ArUco markers are integrated into ORB-SLAM.Thus,the ORB feature points map with real scale is obtained by using the absolute pose estimated by using ArUco markers,and the cumulative error is corrected in the tracking process.Moreover,when the camera can not see the ArUco markers,the absolute pose of the camera in the world coordinate system can still be obtained by tracking the map of ORB feature points.According to the absolute pose information provided by visual odometry,a complete visual navigation system based on ROS is designed,and it is tested on AGV experimental platform.The experimental results show that the mobile robot can accurately reach the target position along the preset navigation path using the pose obtained by the visual odometry,and the navigation process is stable and reliable.