We propose a novel AI-based few-shot self-supervised learning method for automatic optical inspection image quality assessment and component detection based on only few training images. Our method iteratively learns the feature representations of the components by using self-similarity of these components. With the large number of self-learned representations, the appearance variations of each component are then effectively learned in the AI model for component detection and measurement. The computation complexity of our method is significantly lower than that of deep learning methods.
Our method is a novel AI-based few-shot self-supervised learning method for AOI component detection. With only few training images (3 in the experiments), our method can learn a large number of self-learned representations for the appearance variations of each component effectively. In addition, the learned representations are based on the target components which make our method become more general to detect components in testing images. The efficiency of our method makes no use of GPU. It significantly outperforms the state-of-the-art method YOLOv4 in AOI PCB and LCD component detection.
The potential system users are ALL of the manufacturers who need to perform AOI for their product. Our system can be run on Windows with versions above 7 in a general PC WITHOUT GPU, and only requires low memory. The prerequisites for using our system are undemanding. For practical usage, we have evaluated our method on general PCB AOI images for image quality assessment and IC component detection, and achieved more than 95％ accuracy. We have also evaluated our method on AOI LCD images of Array and Color Filter for component detection and measurement, and achieved more than 96％ accuracy.

線上展網址：
https://tievirtual.twtm.com.tw/iframe/c72b95f4-01e6-4e4f-8d30-b31ba5821af6?group=23bfb1fa-dd5b-4836-81a1-4a1809b1bae5&lang=en