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

In the indoor laser simulation localization and mapping (SLAM) system, the signal emitted by the LiDAR sensor is easily affected by lights and objects with low reflectivity during the transmission process, resulting in more noise points in the laser scan. To solve the above problem, this paper proposes a clustering noise reduction method based on keyframe extraction. First, the dimension of a scan is reduced to a histogram, and the histogram is used to extract the keyframes. The scans that do not contain new environmental information are dropped. Secondly, the laser points in the keyframe are divided into different regions by the region segmentation method. Next, the points are separately clustered in different regions and it is attempted to merge the point sets from adjacent regions. This greatly reduces the dimension of clustering. Finally, the obtained clusters are filtered. The sets with the number of laser points lower than the threshold will be dropped as abnormal clusters. Different from the traditional clustering noise reduction method, the technique not only drops some unnecessary scans but also uses a region segmentation method to accelerate clustering. Therefore, it has better real-time performance and denoising effect. Experiments on the MIT dataset show that the method can improve the trajectory accuracy based on dropping a part of the scans and save a lot of time for the SLAM system. It is very friendly to mobile robots with limited computing resources.

In recent years, Simultaneous Localization and Mapping (SLAM) has gradually become a focal point in the field of artificial intelligence applications, such as autonomous driving, and AR/VR, demonstrating its irreplaceable role. TraditionalSLAMis based on assumptions of static scenes, resulting in lowstability and accuracy in dynamic environments. To address this issue, a semantic SLAM system named BiSeNetV2-SLAM, combining ORB-SLAM2, BiSeNetV2, and optical flow fusion, has been proposed. Firstly, the system utilizes the BiSeNetV2 semantic segmentation network to eliminate potential prior dynamic objects. Subsequently, optical flow tracking is employed to detect the motion states of dynamic objects, achieving semantic SLAM in dynamic scenarios. Additionally, the system enhances feature matching accuracy and efficiency by combining the efficient local image descriptor BEBLID with the ORB feature extraction method. In multiple high-dynamic scenes from the TUM public dataset, In multiple high dynamic range scenes from the TUM public dataset, compared to ORB-SLAM2, the RMSE increased by more than 86%, and the S.D. increased by more than 89%. While maintaining similar accuracy compared to DynaSLAM, there was a significant improvement in processing speed.

In visual SLAM (simultaneous localization and mapping) based on conventional point features, factors like rapid perspective changes of camera and environment with weak textures and unclear features can cause the system to fail or even make it impossible to initialize. To solve this problem, we propose a SLAM system with multiple feature points fusion. First, in this system, ORB and GCNv2 features are used in combination to ensure the robustness of tracking threads. Secondly, a new condition for generating keyframes is proposed, which uses relative rotation and translation between frames to determine whether keyframes should be created. Finally, the pose propagation between maps is carried out before the global BA optimization to improve the optimization speed and the reliability of mapping, and make the final map generated by the SLAM system have global consistency. We used the framework of ORB-SLAM3 as a carrier and conducted experiments on Tum datasets and real environments. The experimental results show that the proposed algorithm can complement the advantages of various feature points on the premise of ensuring good accuracy, and effectively improve the robustness of the SLAM system in environments with weak textures, unclear features, and camera’s perspective changes rapidly.