Visual Inertial Navigation Optimization Method Based on Landmark Recognition

This study utilizes a low-cost, low-power, and highly accurate monocular visual-inertial odometry (VIO) as the navigation algorithm for unmanned aerial vehicles (UAV). However, VIO may experience positioning inaccuracies or even loss in cases of lighting changes and insufficient environmental textures. To mitigate the accumulation of errors in the algorithm, global localization information is introduced. In order to address the issue of positioning accuracy in the absence of Global Navigation Satellite Systems (GNSS), this paper proposes an optimized Visual-Inertial Navigation System (VINS) based on landmark recognition, considering the prior information of UAV flight missions. The system relies solely on visual and IMU information, employing the VINS-Mono algorithm for high-precision positioning. After flying a certain distance, the system recognizes landmarks from a landmark database and obtains their prior coordinate information. Finally, the pose graph and optimization algorithm are used to refine the pose of all keyframes. Experimental results indicate that this navigation system effectively reduces the accumulated errors of VINS and enhances the positioning accuracy in scenarios where GNSS signals are completely unavailable.