基于点线特征的鲁棒视觉SLAM系统

urrently, visual simultaneous localization and mapping(SLAM) has received extensive attention as an autonomous localization technology for robots. But the algorithm will fail under the conditions of lighting changes, too fast motion, weak texture environment. In order to improve the robustness of visual SLAM operation, this paper proposes a feature extraction and matching algorithm from coarse to fine point and line based on feature method SLAM. In tracking process, the line feature extraction and matching of part of the image frame is selectively performed according to the matching result of the point feature. In mapping process, line feature extraction and matching are performed on all keyframes newly inserted into the map. Experiments show that the point-line feature visual SLAM system based on this algorithm can meet the real-time tracking requirements, mine more environmental information, create 3D maps with richer structures, and improve the robustness of the visual SLAM system.