Recognition of sticker breakout in mould thermograph based on template matching

Abstract

ABSTRACT Breakout is a serious accident during continuous casting. Sticker is the main type of breakout accidents. The propagation behaviour of sticker breakout in mould is the key to recognizing and preventing breakout accident. A visual monitoring system of breakout has been developed to capture sticker characteristics, which usually consist of hot and cold regions. According to the propagation behaviour of sticker breakouts, four templates were selected with a Euclidean distance algorithm, such as typical, long-span, left one-sided and right one-sided templates. Then, the template matching algorithm was applied to identify the true and false sticker breakouts. The results show that 120 cases of true and false stickers can be distinguished. The recognition rate of true sticker is 100%. There is no false sticker to be determined as a true sticker. This method is not only helpful to capture the propagation behaviour characteristics of sticker breakouts in the mould thermograph but also improves the accuracy of breakout prediction.