Abstract: Aiming at the requirement of autonomous operation of the mobile robot in complex environment,this paper proposes a RGB-D SLAM based on image feature.Compared with the traditional filter method,the error accumulation during the long time motion of the robot is accumulated,adopt SLAM method based on graph optimization.Proposed algorithm is divided into two parts:frontend and backend.The frontend is responsible for processing the image data and extracting the geometric relationship between the poses of the robot,firstly,the feature points of color RGB image are extracted,high dimensional feature descriptor are created and the location correspondence of feature points is established.The backend is responsible for expressing the position and pose of the robot at each moment and diminishing the drift of the trajectory,constructing a graph of the geometric position relationship and its uncertainty,through the optimization of the graph to get the best trajectory,finally,the sparse point cloud map and trajectory are generated.The experimental results show that the proposed method is practical and robust.

Key words: SLAM, RGB-D, graph optimization, feature extraction and matching, loop closure detection