Utilizing Visual Measurements for Obtaining Robust Attitude and Positioning for Pedestrians

Laura Ruotsalainen, Jared Bancroft, Heidi Kuusniemi, Gerard Lachapelle, Ruizhi Chen

Abstract: Pedestrian navigation is mostly needed in environments where GNSS is denied, namely indoors and in urban canyons, in which case augmenting the system with other sensors and information sources is necessary. Other Radio Frequency (RF) systems are widely used for obtaining the position of the pedestrian when GNSS measurements are not available or are degraded, but their drawbacks are the need for an existing infrastructure and the requirement for manual deployment. Self-contained sensors provide infrastructure free relative positioning through attitude and translation. Inertial Navigation Systems (INS) are an excellent option in GNSS denied environments. INS consist of accelerometers providing the translation of the user and gyroscopes for attitude information. Consumer-grade self-contained sensors suffer from excessive drift errors, especially affecting the gyroscope. This ultimately reduces the accuracy of the attitude and consequently the position. Cameras constitute a viable additional sensor that can be used for improving the attitude. An image is a projective transformation of three-dimensional objects into two-dimensional images. Parallel lines in the real world do not appear parallel in images, but seem to intersect at a point, called the vanishing point. When the camera is calibrated, the position changes of the vanishing points may be used to compute the rotation of the camera. When the orientation of the camera with respect to the pedestrian is kept stable, the rotation of the camera provides the attitude of the pedestrian. Erroneous vanishing points lead to incorrect heading measurements and introduce secondary errors when used in an INS. Thus the validity of the vanishing point is evaluated based on the geometry information of the lines used for calculating its position and is similar to a dilution of precision (DOP) value. This paper presents two complementary means of using the attitude information obtained from images to improve the accuracy of an INS-based pedestrian navigation system. Firstly, visual attitude measurements are used as an absolute update to the attitude states within the INS Extended Kalman Filter (EKF). Secondly, the temporal visual attitude (i.e. change in attitude between consecutive images) is used as attitude updates to the EKF. The improvement of the attitude accuracy when augmented with visual attitude measurements is evaluated. The performances of the two approaches are compared. The visual-aiding is shown to significantly improve pedestrian attitude accuracy in the indoor test scenarios presented by 93 % and 40 % for the first and second methods, respectively.
Published in: Proceedings of the 25th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2012)
September 17 - 21, 2012
Nashville Convention Center, Nashville, Tennessee
Nashville, TN
Pages: 2454 - 2461
Cite this article: Ruotsalainen, Laura, Bancroft, Jared, Kuusniemi, Heidi, Lachapelle, Gerard, Chen, Ruizhi, "Utilizing Visual Measurements for Obtaining Robust Attitude and Positioning for Pedestrians," Proceedings of the 25th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2012), Nashville, TN, September 2012, pp. 2454-2461.
Full Paper: ION Members/Non-Members: 1 Download Credit
Sign In