Ilchuk M.
PhD Student
Lviv Polytechnic National University, Ukraine
Bandery, 12, Lviv, Ukraine, 79013
Stadnyk A.
PhD Student
Lviv Polytechnic National University, Ukraine
Bandery, 12, Lviv, Ukraine, 79013
Prokhorenko S.
Prof., Doctor of Technical Sciences
Lviv Polytechnic National University, Ukraine
Bandery, 12, Lviv, Ukraine, 79013
Abstract
The integration of automated quality control technologies is nowadays a fundamental requirement for modern PCB production. This automated approach replaces manually based methods that are error-prone with fast, accurate and measurable estimates, helping to reduce the time and costs associated with traditional testing. Human inspection is less efficient, slower and overpriced than automated inspection. In this context, digital image processing, especially for the detection of defects such as faulty components or missing elements, becomes crucial. The production of printed circuit boards requires strict quality control to ensure the reliability and efficiency of electronic devices. Improving the detection and analysis of defects on printed circuit boards is as critical as ever. In addition to this, we propose the concept that combining classical computer vision techniques, such as edge detection and Kenny contour analysis, with large language models opens up innovative possibilities in defect identification.
The main function of the system is to comprehensively detect defects on printed circuit boards using automated visual inspection. It uses a digital camera to capture images of each PCB, which are then processed by a computer. This processing includes converting to greyscale and binary shapes, followed by an XOR operation to extract the necessary information. Contour analysis is then applied for classification. Incorporating language models into defect documentation and description is a significant step in automating quality control, as they can not only describe defects but also suggest corrective measures.
This system is capable of detecting various PCB defects such as missing components, incorrect polarity, open circuits and missing tracks. Its advanced fault detection and classification increase the speed and accuracy of evaluation, reducing human error, which is prevalent in quality testing, and introduces a usable fault and defect notification, detailed description and reporting system. Replacing manual testing methods with this new model significantly increases productivity.
Keywords: Image Processing, Printed Circuit Board, Defect Detection, Edge Detection, Classification system, Large language system.