The classifying algorithms for train barriers recognition on the basis of image fusion methods and neural networks

This paper presents the results of new research on neural network classifiers for barriers on traffic routes of locomotives. The general structure of the system of detection, recognition of traffic barriers, input information coming from the subsystem of the computer vision and the subsystem that scans spaces in front of the locomotive are considered. The methods of image fusion based on an adaptive processing of wavelet coefficients are used to increase the resolution of images. A multilayer neural network with error back propagation is used as a classifier.