Space and Picture Models for Machine Vision in Railway.

Abstract

In this paper, space and picture models for machine vision in railway were developed. The railway space is modeled as an imaginary rectangular tunnel including the construction gauge. The walls of the imaginary tunnel are mapped with wall pictures representing wayside facilities and a background landscape. The wall pictures are quasi-normalized along the railway axis. Using the model, we can transform a point in a wall picture into a point in a perspective picture taken by a camera. Consequently, the wall picture and the perspective picture can be transformed, each from the other. We can estimate the mapping error corresponding to deviations between the actual point and the wall. Additionally, some applications showing the effects of this model are illustrated.