Inspection of rail surface defect based on machine vision system

Abstract

An inspection system based machine vision technique for rail surface defects is presented. Automatic detection of noised cracks on surface is analyzed by means of image processing. Image enhancement, denoising and feature extraction are applied in processing the images of track. The region of defect is extracted on the adaptive threshold. Based on morphology of cracks, dynamic template is designed to detect continuous crack boundary, and then length of cracks can be estimated.