Multi-features visual odometry for indoor mapping of UAV

In the near-decade, Visual SLAM (Simultaneous Localization and Mapping) system is becoming more and more important for the navigation of unmanned aerial vehicle (UAV) system, because it is effective to replace positioning devices such as GPS (Global Position System) in the indoor scenes. However, there are still challenges when the camera working in the low-texture scenes. The visual SLAM algorithm based on a single feature is difficult to obtain enough features. The positioning accuracy and robustness of the whole system will be reduced or even cannot work properly. Targeting at these issues, we present the MF-VO: an optimization-based multi-features visual odometry. The algorithm extracts both the point feature and the line feature in the image frame. It enhances the robustness of the SLAM system in low-texture scenes and makes the system more robust. A more accurate visual pose can be obtained by minimizing the reprojection error of local points and lines. Additionally, our algorithm is validated on the EuRoC MAV datasets. MF-VO has proven to be effective in improving UAV robustness and accuracy compared to another advanced visual odometry method.