An improved adaptive ORB-SLAM method for monocular vision robot under dynamic environments

The vision-based simultaneous localization and mapping (SLAM) method is a hot spot in the robotic research field, and Oriented FAST and Rotated BRIEF (ORB) SLAM algorithm is one of the most effective methods. However, there are some problems of the general ORB-SLAM algorithm in the dynamic environment, which need to be solved further, including the control of the number of the feature points and the treatment of the dynamic objects. In this paper, an improved ORB-SLAM method is proposed for the monocular vision robot under dynamic environments. In the proposed method, a concept of reliability is proposed to mark the feature points, which can control the number of the feature points dynamically into the preset range. Then an improved frame difference method based on the partial detection strategy is used to detect the dynamic objects in the environment. In addition, a novel treatment mechanism for the tracking failure issue is introduced into the ORB-SLAM algorithm, which can improve the accuracy of the localization and mapping. Finally, the experiments on the public datasets and private datasets are conducted and the results show that the proposed method is effective.