Visual Localization of a Ground Vehicle Using a Monocamera and Geodesic-Bound Road Signs

Abstract

The work explores the on-board navigation system of the moving object, which includes a camera, odometer and steering sensor. The navigation algorithm is based on the method of visual odometry compensated by the data of recognized road signs. Recognition of road signs is carried out on the basis of a roll-out neural network. The algorithm is tested on natural data.