Automated inspection of solder joints-a neural network approach

摘要

This paper describes a PC-based system for automated inspection ofsolder joints using neural networks. It presents extensive applicationof neural networks to solder joint quality data in the form of visualimages. Numerous methods for data compression and feature extractionhave been applied to enhance the performance of the neural networks. Upto 92 per cent accuracy in identifying solder joint defects was achievedusing visual images. This discussion deals with visible light imagesonly but all techniques may be extended equally to X-ray laminographicimages as preliminary results from such applications indicate