Date of Award
Master of Science in Electrical Engineering
Department of Electrical and Computer Engineering
John F. Raquet, PhD.
The purpose of this research was to obtain a navigation solution that used real data, in a degraded or denied global positioning system (GPS) environment, from low cost commercial o the shelf sensors. The sensors that were integrated together were a commercial inertial measurement unit (IMU), monocular camera computer vision algorithm, and GPS. Furthermore, the monocular camera computer vision algorithm had to be robust enough to handle any camera orientation that was presented to it. This research develops a visual odometry 2-D zero velocity measurement that is derived by both the features points that are extracted from a monocular camera and the rotation values given by an IMU. By presenting measurements as a 2-D zero velocity measurements, errors associated with scale, which is unobservable by a monocular camera, can be removed from the measurements. The 2-D zero velocity measurements are represented as two normalized velocity vectors that are orthogonal to the vehicle's direction of travel, and are used to determine the error in the INS's measured velocity vector. This error is produced by knowing which directions the vehicle is not moving, given by the 2-D zero velocity measurements, in and comparing it to the direction of travel the vehicle is thought to be moving in. The performance was evaluated by comparing results that were obtained when different sensor pairings of a commercial IMU, GPS, and monocular computer vision algorithm were used to obtain the vehicle's trajectory. Three separate monocular cameras, that each pointed in a different directions, were tested independently. Finally, the solutions provided by the GPS were degraded (i.e., the number of satellites available from the GPS were limited) to determine the e effectiveness of adding a monocular computer vision algorithm to a system operating with a degraded GPS solution.
DTIC Accession Number
Rohde, Johnathan L., "Urban Environment Navigation with Real-Time Data Utilizing Computer Vision, Inertial, and GPS Sensors" (2015). Theses and Dissertations. 56.