A Real-Time Contrasts Method for Monitoring Image Data

Machine vision inspection integrated with statistical process control is increasingly being used to detect the surface defects in industrial products. The existing methods utilized traditional control charts to monitor the change point or abnormal regions of images. However, there are some challenges in handling high-dimensional and complex image data. This paper proposes to monitor image data based on the real-time contrasts (RTC) method. RTC method converts the monitoring problem to a continuous classification problem by labelling the reference images and the real-time images with different labels. Classification accuracy is used to build statistics for image data monitoring. And a kind of under-sampling method is used to tackle the problem of sample imbalance, which exists in process and has negative effects on the classification accuracy. Performance of the proposed method is compared with that of an alternative method under both single and multiple defects scenarios via simulations. The results show that the proposed method performs better in some situations and can effectively identify the occurrence of shifts in the process.