SLAM Fusion Optimization Based on Monocular Vision and Inertial Sensor

Simultaneous Localization and Mapping (SLAM) is a key technology in the field of robotics, and it is the key part and foundation for the complete autonomous control and intelligence of robots. With the continuous development of 3D computer vision algorithms, the vision-based SLAM method had become a research hotspot in recent years. However, the visual SLAM method relies too much on the feature information of the surrounding environment, cannot deal with the scene texture missing and dynamic scenes. Moreover, the visual sensor frame rate is low, which can not handle the fast motion of the target. The Inertial Measurement Unit (IMU) measures the angular velocity and acceleration of the sensor itself, which has significant complementarity with the vision sensor, and has the potential to build a more robust SLAM system. In this paper, a fusion optimization method is proposed, in which the local optimization and global optimization methods are used in parallel. The local optimization uses the optimization of sliding window to obtain pose information. The global optimization focuses on the information of all points and gives a method to correct each of the previous key frame. This paper also proposes a relative marginalization method to ensure the global consistency of optimization and improve the precision of optimization. The experimental results show that the proposed method can obtain accurate target motion trajectory and improve the positioning accuracy of SLAM.