Visual detection based on computer vision for sticker breakout in slab continuous casting

Abstract

Based on the temperature readings of the mould measured by thermocouples during slab continuous casting, mould temperature and its velocity thermographs were built up, and a visual detection method for the sticker breakout proposed. In the temperature velocity thermograph, the abnormal temperature regions are marked by virtue of the computer image-processing algorithm, such as frame difference, threshold segmentation and eight connected component labelling. The abnormal region features are then extracted and processed as the temperature velocities, geometry characteristics and propagation velocities. In order to distinguish the true and false sticker breakouts, a comparison analysis is made. The results show that the combination of the temperature velocity thermograph and computer vision algorithm is an effective method for the sticker breakout detection. The temperature velocities, geometry characteristics and propagation velocities of the abnormal regions are three important criteria for sticker breakout detection. This method also provides a technology for developing a more visual and intelligent mould monitoring system and is helpful to enhance the accuracy for the breakout prediction system.

Keywords:

• Continuous casting
• Visual detection
• Sticker breakout
• Image processing
• Feature extraction