Construction of pattern recognition system optimized for X-ray inspection of plastic electronics and OLED displays

摘要

It is well known that the quality of plastic electronics, printed circuit boards (PCB), and Organic Light Emitting Diode (OLED) displays can be variable in production. This in turn may lead to an increasing number of faulty batches. Digital radiography and pattern recognition allow inspection of the products non-destructively in real time to increase the quality of output batch for manufacturing lines. An important part of digital radiography for non-destructive testing (NDT) is image processing and pattern recognition. In this study, the whole chain of data processing is reconsidered starting from building representative reference data sets, image quality assessment, selecting region of interest, noise reduction, contrast enhancement, image segmentation. It is expected to consider different pattern recognition techniques such as active shape model, moving windows and greedy pursuit as well as traditional cross-correlation and distance template matching.