A Real-Time Map Restoration Algorithm Based on ORB-SLAM3

In the monocular visual-inertia mode of ORB-SLAM3, insufficient excitation obtained by inertial measurement unit (IMU) will lead to a long system initialization time. Hence, trajectory can be easily lost and map creation not completed. To solve this problem, fast restoration method is proposed in paper, which adresses problem IMU. Firstly, frames before are quickly tracked using bag-of-words maximum likelyhood perspective-n-point (MLPNP). Then, grayscale histogram used accelerate loop closure detection reduce time consumption caused restoration. After experimental verification on public datasets, algorithm establish complete ensure real-time performance. Compared with traditional accuracy improved about 47.51% efficiency 55.96%.