Multi-Structural Instrument for Identifying Surface Defects on Rails

Abstract

A multi-structural instrument for identifying surface defects on rails has been developed. The instrument operates based on intelligent analysis of video data, and it opens up a wide range of possibilities for using neural-network analysis of images in real time. The instrument can transform a color image into a zero-contrast image, normalize video images, and convert them to binary form. It is distinguished by its real-time noise suppression, evaluation of indicators that provide information on defects, and automatic neural-network classification of defects. The instrument is used together with a dynamic expert system that employs a production model of knowledge representation.