A Visual-Inertial Navigation System Based on Multi-State Constraint Kalman Filter

摘要

Currently, fusion algorithms for visual and inertia data have become a research hotspot in the field of computer vision and robotics. Compared to loosely coupled method, the tightly coupled one can obtain higher accuracy and robustness. In this paper, a tightly coupled visual-inertial system is implemented based on Multi-State Constraint Kalman Filter (MSCKF) [1] considering both accuracy and efficiency in order to navigate a micro air vehicle (MAV) in GPS-denied indoor environment. Firstly, a C++ implementation of the standard MSCKF with a feature matching front-end is realized. Then a new feature tracking front-end based on modified optical flow is proposed. Experiments show that the new method is faster than the feature matching method and has almost the same accuracy. Finally, a Hybrid-MSCKF algorithm is implemented to reduce the drift. The algorithm is tested on a MAV dataset, and experiments show that the robustness and positioning accuracy of the algorithm reach navigation requirements of indoor MAVs.