Integration of the inertial navigation system with consecutive images of a camera by relative position and attitude updating

摘要

This paper introduces a new method for improving the inertial navigation system errors using information provided by the camera. An unscented Kalman filter is used for integrating the inertial measurement unit data with the features' constraints extracted from the camera's image. The constraints, in our approach, comprise epipolar geometry of two consecutive images with more than 65% coverage. Tracking down a known feature in two consecutive images results in emergence of stochastic epipolar constraint. It emerges in the form of an implicit measurement equation of the Kalman filter. Correctly matching features of the two images is necessary for reducing the navigation system errors because they act as external information for the inertial navigation system. A new method has been presented in this study based on the covariance analysis of the matched feature rays' intersection points on the ground, which sieves the false matched features. Then, the inertial navigation system and matched feature information is integrated through the unscented Kalman filter filter and the states of the vehicle (attitude, position, and velocity) are corrected according to the last image. In this paper, the relative navigation parameters against the absolute one have been corrected. To avoid increasing dimensions of the covariance matrix, sequential updating procedure is used in the measurement equation. The simulation results show good performance of the proposed algorithm, which can be easily utilized for real flights.