| Hot-rolled strip surface inspection in the steel industry is of great significance in theproduct quality control. It is developed to reduce defect rates and scrap defective, and toimprove the competitiveness of the enterprises themselves. Recently computer vision andimage processing techniques are adopted to achieve surface defect detection on the stripedsteel line. It is necessary to collected a large number of images of defective strips and useadvanced image processing techniques, in order to provide real-time processing for theaccurate product quality control.Image processing technology is maturing in recent years, its ability and accuracy hasbeen greatly improved. However, due to the limitation of industrial site testing environmentand the hardware configuration of the device, there are always some false results of the stripdetection caused by the remaining water on the steel surface. These false detecting results leadto the detected piece of steel with their images not treated. These false alarms bring a greatwaste of the hardware, and ultimately affect the normal striped steel production.In this thesis a striped steel surface defect system using image processing techniquesbased on OpenCV is proposed. Recent technological achievements have been reviewed.Through detailed analysis of the structure of the original system and its detection process, thecauses for the impact of the monitoring system are identified. Both hardware and softwareimprovements have been made. In hardware, the original storage drives have been extended to1T and the strip surface purging apparatus are added on the line; In software, a cornerdetection method is developed to determine whether the striped steel surface is covered withthe unevaporated water. With the correct classification of the defect image caused by water,some false alarms are eliminated, and the system availability and reliability are greatlyimproved.Finally, improvement aspects and outlooks of the system are summarized. |
PDF Full Text Request