Tightly Coupled Stereo Vision Aided Inertial Navigation Using Continuously Tracked Features for Land Vehicles

Fei Liu, Yashar Balazadegan Sarvrood and Yang Gao

Peer Reviewed

Abstract:	Visual odomestry with monocular or stereo cameras becomes an effective way of navigation by utilizing the continuously tracked features (land marks) in consecutive frames to estimate the position, velocity and attitude of camera. There is inevitable drift of visual odometry with error accumulated with time. People explore integration of cameras with other sensors to extend the duration of providing reasonable solutions. INS/Vision integration has gained increasing attention to reduce the error accumulation of both sensors. The typical way of integration is to process two kinds of sensors independently and fuse the processing results. In this paper, we proposed a tightly coupled method of INS and stereo cameras with the direct pixel coordinates obtained by camera as measurements. The formulation of proposed method is derived and it is tested with the outdoor datasets provided by KITTI (Karlsruhe Institute of Technology and Toyota technological Institute). The results show better performance of the proposed method, compared with classical visual odometry approach.
Published in:	Proceedings of the 28th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2015) September 14 - 18, 2015 Tampa Convention Center Tampa, Florida
Pages:	2127 - 2133
Cite this article:	Liu, Fei, Sarvrood, Yashar Balazadegan, Gao, Yang, "Tightly Coupled Stereo Vision Aided Inertial Navigation Using Continuously Tracked Features for Land Vehicles," Proceedings of the 28th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2015), Tampa, Florida, September 2015, pp. 2127-2133.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In