Automatic Identification and Classification of Surface Defects in Small-Sized Optics
Fanshu Zhang,Dingrong Yi,Linghua Kong,Jishi Zheng,Tao Yuan,Siling Chen
DOI: https://doi.org/10.1088/1361-6501/ad6baf
IF: 2.398
2024-08-08
Measurement Science and Technology
Abstract:Precision optics are widely used in lighting systems, imaging systems and high precision inspection equipment. The presence of surface defects in optical components can seriously affect the design performance of the equipment, so the optical components must be rigorously inspected to eliminate defective optical components. Currently, engineering products are moving toward smaller sizes, resulting in smaller optical elements. However, realizing small-sized optical elements is a great challenge because it requires micrometer-scale high-resolution image acquisition and automatic detection of defects on their surfaces. A machine learning-based surface defect detection method is proposed for a wide variety of surface defects and small sizes. A dark-field micro-scattering imaging system is used to acquire dark-field images. Then image segmentation and feature analysis are performed on each image to extract the original feature data and compressed with relifF algorithm. Based on the compressed feature data, a classification model is built using a support vector mechine(SVM). The detection method can classify three kinds of defects, namely, dig, scratch, and scuff mark, with an accuracy of up to 98%. The experimental results of this study show that the method can effectively and accurately detect several microdefects occurring in small-size optical components, thus providing valuable insights for the optimization of mass production parameters of optical components.
engineering, multidisciplinary,instruments & instrumentation