基于先验特征地图匹配约束的视觉/惯性定位方法

Monocular visual-inertial positioning system is widely used due to its low cost and high autonomy. However, visual-inertial odometry generates large cumulative errors after a long period of operation, which cannot provide accurate pose estimation, and there are feature point tracking failures in the system with complex light and road conditions, which cannot solve the current position with low stability. To address the above problems, a priori feature map matching method based on nonlinear optimization is proposed. The pose graph established by monocular visual-inertial SLAM is used to constrain the residual estimation of the system. By calculating the similarity between the current frame and the historical frames in the feature map, and using the bit-pose of the map for error constraint if the similarity between them is higher than a certain threshold, the cumulative error is eliminated and the stability of the system is improved. The map matching method is verified using data from an UAV flying at an altitude of about 6.4 km for 15 minutes. The experimental results show that the proposed map matching algorithm improves the navigation accuracy (RMSE) by 91% in the case where the time-consuming increase is less than 20% compared with the unused map matching algorithm, and can effectively reduce the cumulative errors of monocular vision-inertial system and improve the stability of the system.