Research on SLAM System Based on Binocular Vision and IMU Information

Abstract

Simultaneous localization and mapping (SLAM) is one of the key technologies in the field of robotics, is the key to autonomous navigation of mobile robots, and is also the core and foundation of autonomous and intelligent mobile robots. The SLAM method which relied information solely on monocular cameras is too dependent on the characteristic information of the surrounding environment. For the lack of the scene texture, the dramatic changes of the illumination and the poor performance of the dynamic scene, and the low frame rate of the visual sensor, it cannot deal with the situation of fast motion. The inertial measurement unit (IMU) can output the acceleration and angular velocity of the sensor itself at a high frame rate, and is not affected by the environment, but with serious drift. To solve this problem, this paper aims to design a SLAM system which combines binocular vision information and IMU information. It can realize robust and precise positioning in unknown environment, and provide corresponding navigation map for navigation.