Semi-direct RGB-D slam algorithm for mobile robot In dynamic indoor environments

Abstract

In order to solve the problem of accurate navigation of mobile robots in dynamic indoor environment, a semi-direct RGB-D visual SLAM (Simultaneous Localization and Mapping) algorithm based on motion detection algorithm is proposed. The algorithm is mainly divided into three parts: motion detection, camera positioning and dense map construction based on TSDF (Truncated Signature Distance Function) model. Firstly, a preliminary estimation of the pose of the camera is achieved by using a sparse image alignment algorithm. Then, a real-time updated Gaussian model based on image patches is established to segment moving objects in the image. Based on this, the local map points projected in the moving area of image are eliminated, and the pose of the camera is further optimized. Finally, the TSDF dense map is constructed by using camera pose and RGB-D camera image information. The dynamic update of the map in real time is achieved by using the image motion detection result and the color change of the map Voxel. The experimental data under TUM dataset show that the proposed algorithm can effectively improve the camera positioning accuracy and real-time update dense map in indoor dynamic environment, which greatly enhances system robustness and environmental information for robot sensing.