Vision-based inspection for periodic defects in steel wire rod production

摘要

Abstract. Defective rollers in rolling plants often create spatially periodicndefects on the product surface. It is essential for operators to receivenearly warnings of such periodic defects and take appropriate action tonavoid a large quantity of defective products. We present a robust andnefficient real-time method for inspecting periodic defects in continuousnsteel wire rods. The proposed inspection method consists of two parts:nu00011u0002 an algorithm for detecting all defect candidates and u00012u0002 an algorithmnto detect periodicities in the detected defect candidates and to determinentheir periods. The first algorithm exploits the translation-invariant prop-nerty of the Haar undecimated discrete wavelet transform to improve thensignal-to-noise ratio of the surface image. The second algorithm is basednon analysis of the frequency spectrum of defect candidate positions. Itnidentifies periodicities of the defect candidates by finding repeatingnimpulse-like spectra. The inspection system using the proposed methodnis applied to a real production line and detects periodic defects in realntime and identifies their periodicities with high accuracy. © 2010 Society ofnPhoto-Optical Instrumentation Engineers. u0003DOI: 10.1117/1.3284779