A Low-Texture Monocular Visual Odometer Based on Point-Line Feature

When the camera is in a low-textured scene, it is difficult for the simultaneous localization and mapping (SLAM) algorithm based on point features to track sufficient effective point features, which leads to low accuracy and weak robustness, or even results in algorithm failure. To tackle this problem, a low-texture monocular visual odometer based on point-line feature is proposed in this work. Firstly, the data association accuracy and robustness are improved by using the complementation of point-feature and line-feature. Based on this, an environmental-feature map with geometric information is constructed. We thoroughly evaluate this approach and the new initialization strategy on the TUM RGB-D benchmark. The results demonstrate that the use of lines does not only improve the original ORB-SLAM solution in poorly textured frames, but also systematically improve the solution in sequence frames combining points and lines without compromising the efficiency.