Real-Time RGB-D SLAM with Points and Lines

SLAM is regarded as one of the core technology for automatic robots and self-driving cars, which has proven its value and gain much attention recent years. However, challenges still exist and problems such as changeable illumination and low-textured environment make SLAM algorithms unavailable, especially for feature-based visual SLAM. In this paper, a real-time RGB-D SLAM with points and lines is proposed, which has increased the accuracy and availability of SLAM algorithm in such situations. Specifically, our work is built upon ORB-SLAM, which is currently the state-of-the-art visual SLAM algorithm. The extracting method of point features in it is modified to reduce the influence of illumination changing. And instead of extracting only points, lines are extracted simultaneously for better localization. Furthermore, the idea of stability toward points and lines is proposed, and its measurement will be used in the bundle adjustment. Finally, the experiment shows that our approach has better accuracy and availability compared to the original ORB-SLAM.