Multi-illumination Imaging System with CNN Machine Learning Applied to Optical Inspection of Highly Complex Specimen

This paper develops a method to improve the image recognition ability of automated optical inspection (AOI) system. Through multi-directional illumination, the optical sensor can capture a large number of different optical images. These images become the learning materials of the convolutional neural network (CNN), and then gradually improve the recognition ability of the AI software. The experimental specimen is stool parasite cysts. This kind of sample still relies on the human eye for judgment. Because traditional automated optical inspection systems are limited by their specificity, it is difficult to identify samples with overly complex backgrounds. Experimental results show that the developed method can significantly improve the accuracy of image recognition. The purpose of this study is to develop a new method to improve automated optical inspection technology and its application.