Loosely-Coupled Stereo Vision-Aided 3D Reduced Inertial Sensor and GPS for Land Vehicle Localization

Yashar Balazadegan Sarvrood and Yang Gao

Peer Reviewed

Abstract:	Monocular and binocular visual odometry has attracted a lot of attentions due to their advantages such as low cost and rich texture recently. Visual odometry is an effective dead reckoning system which estimates the 6 degrees of freedom (6DoF) ego-motion including three position increments and three alignment increments by using the continuously tracked features in consecutive frames. Like any other dead reckoning system, visual odometer accumulates errors over time. Integration of camera with a wide variety of sensors such as odometer, laser scanner, IMU (Inertial Measurement Unit) and GPS reduces the unavoidable visual odometry error growth rate and extends the duration of reasonable navigation solution. IMU and camera are complementary sensors and the integration of these two sensors reduces the error accumulation of both sensors. Several loosely and tightly coupled integration of IMU and vision have been developed during the last decade which uses cameras and full IMU including three accelerometers and three gyroscopes. In this paper, we proposed a new method for reduced IMU (two horizontal accelerometers and one vertical gyro) and vision integration called RISAVS (Reduced Inertial Sensor And Vision System). RISAVS eliminates several error sources that exist when using a traditional full IMU (especially the low cost MEMS (Micro Electro Mechanical Sensor) based IMUs) in land vehicles. RISAVS method formulation was derived and it was tested with online datasets provided by KITTI (Karlsruhe Institute of Technology and Toyota technological Institute). The proposed method performed better than classical visual odometry. Furthermore, it reduces the cost of the system by using reduced IMU instead of full IMU.
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:	2152 - 2160
Cite this article:	Sarvrood, Yashar Balazadegan, Gao, Yang, "Loosely-Coupled Stereo Vision-Aided 3D Reduced Inertial Sensor and GPS for Land Vehicle Localization," Proceedings of the 28th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2015), Tampa, Florida, September 2015, pp. 2152-2160.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In