Visual Relocalization Optimization of Mobile Robot via Combining AprilTag with ORBSLAM3

In dynamic environment, how to achieve fast and accurate relocalization when the current pose information is lost has always been a difficult task in visual simultaneous localization and mapping of mobile robot. The inconsistency of the scene in which the mobile robot is located due to dynamic objects can make it impossible to match the current image with the a priori map. Thus, increasing the relocation error and even causing the relocation to fail. To solve the above situation, we propose a visual relocalization optimisation mechanism that combines AprilTag artificial beacons with ORBSLAM3. By using AprilTag as one of the bases for selecting candidate keyframes for relocalization in the ORBSLAM3 system, a hybrid feature candidate keyframe judgement is established, and a hybrid feature pose optimisation algorithm is proposed based on the high-precision 6 DOF pose information of AprilTag. The above algorithm is used to improve the robustness and accuracy of the mobile robot's relocalization when it loses its own pose estimation. The experimental results show that the proposed method can effectively improve the relocalization accuracy and robustness of the mobile robot in dynamic scenes with large scene changes, difficult relocalization and inaccurate relocalization.