Inertial Navigation Aiding Using Multiple View Feature Matching

Tsahi Aviv
Work towards M.Sc. degree under the supervision of Prof. P. Gurfil and Dr. H. Rotstein
Faculty of Aerospace Engineering
Technion – Israel Institute of Technology

Inertial navigation is ubiquitous; however, it tends diverge due to imperfections in the inertial sensors, resulting in growing position and attitude errors. Thus, it is a common practice to aid the inertial navigation using external sensors. In the current research, we develop a method for aiding the inertial navigation using optical information captured by an onboard wide field-of-view (FOV) camera.

The basic concept is to use a sequence of frames captured by the camera along the platform trajectory. Once detecting the same features in a number of sequential frames, it is possible to use the feature motion across the detector matrix as the estimation innovation. A key advantage of this concept is the fact there is no need of any a priori information regarding the scene.

Due to the problem nonlinearity, there are no existing closed-form solutions for the aiding process, and the common estimators used are the extended Kalman filter (EKF) or bundle adjustment, which both require restrictive assumptions. In this research, we solve the aiding problem from a maximum likelihood point of view, and develop an optimal estimator under a Gaussian assumption, which enables to fuse inertial measurements with the multiple feature optical measurements. The performance of the new method is evaluated in a number of realistic scenarios.

The talk will be given in Hebrew

Wed, 18-06-2014, 16:30 (Gathering at 16:00)

Classroom, ground floor, Library, Aerospace Eng.

Light refreshments will be served before the lecture

Inertial Navigation Aiding Using Multiple View Feature Matching