Dynamic RGB-D SLAM Based on Static Probability and Observation Number

This article proposes a simultaneous localization and mapping (SLAM) method for RGB-D cameras in dynamic scenes, which can effectively overcome the influence of dynamic objects and improve the pose estimation accuracy of RGB-D cameras in dynamic scenes. To detect the feature points on the dynamic objects, a dynamic feature point detection method is proposed, which is based on double K-means clustering. Then, a static weight is established for each feature point in the current frame to represent the probability that the feature point is static. The static weight is composed of static probability and static observation number (SON). Finally, to make the traditional random sample consensus (RANSAC) algorithm more suitable for dynamic scenes, the proposed method improves the RANSAC algorithm. Experimental results show that the proposed method can effectively improve the pose estimation accuracy of the RGB-D camera in dynamic scenes.