The rise of smart devices manufacturing brings out the problem of glass inspection. When performed by humans, this task is costly, time-consuming, and inconsistent. Therefore, a recent study suggests an intelligent localization and classification of tiny defects based on semi-supervised learning.

” data-src=”https://www.technology.org/texorgwp/wp-content/uploads/2020/05/smir-720×480.jpg” alt=”” width=”584″ height=”389″ data-srcset=”https://www.technology.org/texorgwp/wp-content/uploads/2020/05/smir-720×480.jpg 720w, https://www.technology.org/texorgwp/wp-content/uploads/2020/05/smir-400×267.jpg 400w, https://www.technology.org/texorgwp/wp-content/uploads/2020/05/smir-768×512.jpg 768w, https://www.technology.org/texorgwp/wp-content/uploads/2020/05/smir-450×300.jpg 450w, https://www.technology.org/texorgwp/wp-content/uploads/2020/05/smir.jpg 1280w” data-sizes=”(max-width: 584px) 100vw, 584px”>

Image credit: Victorgrigas, Wikimedia (CC BY-SA 3.0)

It can work with full smartphone glass images without cropping the transparent region from it. The method can classify detects into scratches, pits, and light leakage and differentiate them from sensor regions or light reflections due to dust.

The method consists of four stages: suspicious regions detection, feature extraction using a pre-trained convolutional neural network, recognizing non-defects using a background/defects classifier, and a random-forest-based six-class defects classifier. The defects which cannot be seen by the human eye (up to 5 microns) are detected so the technology can outperform manual inspection.

The presence of any type of defect on the glass screen of smart devices has a great impact on their quality. We present a robust semi-supervised learning framework for intelligent micro-scaled localization and classification of defects on a 16K pixel image of smartphone glass. Our model features the efficient recognition and labeling of three types of defects: scratches, light leakage due to cracks, and pits. Our method also differentiates between the defects and light reflections due to dust particles and sensor regions, which are classified as non-defect areas. We use a partially labeled dataset to achieve high robustness and excellent classification of defect and non-defect areas as compared to principal components analysis (PCA), multi-resolution and information-fusion-based algorithms. In addition, we incorporated two classifiers at different stages of our inspection framework for labeling and refining the unlabeled defects. We successfully enhanced the inspection depth-limit up to 5 microns. The experimental results show that our method outperforms manual inspection in testing the quality of glass screen samples by identifying defects on samples that have been marked as good by human inspection.

Link: https://arxiv.org/abs/2010.00741


You can offer your link to a page which is relevant to the topic of this post.

Leave a Reply

Your email address will not be published. Required fields are marked *