Detection of Defects in PCB Images by Numbering, Measurement of Chain Features and Machine Learning

The approach contains algorithms for determining connection and resistance defects. They are thinning, numbering, comparison of corteges, and measurement of the trace resistance. Imposing a tolerance on the concentrated resistance changes it is possible to mark the suspicious printed circuit boards and calculate data for application of Machine Learning. All software procedures solve the task of data preparation for multi-layer neural networks Brain.js which divides printed circuit boards into two classes: defective and working.