In the field of advanced optical manufacturing and ultra-precision machining, achieving digital quantitative inspection of surface defects (such as scratches, pits, and bubbles) is crucial for ensuring the sustainable development of technology. As the manufacturing industry increasingly prioritizes product quality and competitiveness, the widespread adoption of machine vision technology for automated inspection has become an inevitable trend.

The surface quality of precision components is typically evaluated through parameters like surface form, roughness, and surface defects. However, unlike the former two, for which mature commercial inspection instruments (such as interferometers and profilometers) already exist, standardized digital inspection equipment for surface defects remains a gap in the industry.

Despite significant advancements in optical manufacturing technology during the 20th century, it wasn't until the early 21st century that researchers gradually recognized that once component surface form and roughness were well-controlled, surface defects were increasingly becoming the primary bottleneck limiting process levels and performance improvement. This is particularly true in fields demanding extremely high optical precision, such as inertial confinement fusion, very large-scale integrated circuits, and high-end manufacturing, where any isolated sub-micrometer microscopic defect can lead to catastrophic system failure. These very fields are critical indicators of a nation's comprehensive strength.

The research by Professor Yang Yongying's team at Zhejiang University originated from the challenge of quantitatively evaluating surface defects on large-aperture, ultra-smooth components within inertial confinement fusion systems. This requires detecting sub-micrometer microscopic flaws on macroscopic surfaces up to hundreds of millimeters in size. To address this, the team has dedicated over two decades to utilizing machine vision methods for the quantitative evaluation of scratch and pit-like defects.

Distinct from typical machine vision books that emphasize image processing algorithms, this work focuses on the cutting-edge field of digital quantitative inspection of optical surface defects, systematically elucidating the following innovative achievements:

I. Theoretical Breakthroughs

Established a microscopic scattering dark-field imaging model based on the Finite-Difference Time-Domain (FDTD) method, enabling electromagnetic scattering simulation of sub-micrometer defects under strong illumination.

Pioneered a reverse identification database for near/far-field scattered light of surface defects, providing a theoretical basis for inverse calibration of microscopic defects.

II. Key Technologies

Proposed ray tracing imaging simulation theory and radiometric scattering characterization methods, replacing traditional experimental debugging to achieve parameterized virtual modeling of inspection systems.

Overcame the challenge of defect detection on high-order aspheric/freeform surfaces: Achieved full-aperture 3D digital defect evaluation through optimal illumination layout, automatic centering modeling, and projection transform image stitching techniques.

III. Industrial Applications

Developed deep learning-based defect recognition algorithms (object recognition networks, semantic segmentation networks, class-imbalanced semi-supervised learning) to meet the demands of online inspection for glass panels and complex assemblies.

Led the formulation of national standard GB/T 41805-2022 (Microscopic Scattering Dark-Field Imaging Method), filling the gap in standardized surface defect inspection.

Professor Yang Yongying's team at Zhejiang University has deep expertise in the field of digital quantitative inspection of surface defects for many years, focusing on the quantitative evaluation of optical surface defect imaging. They discuss international standards, US standards, and Chinese national standards for defect evaluation, and have established a novel system for digital quantitative evaluation of surface defects.

'Optical Imaging and Defect Evaluation in Machine Vision' is a core work published by Tsinghua University Press as part of the '14th Five-Year Plan' National Major Publishing Project, 'Revolutionary Optical Science and Technology Series.' It provides theoretical tools and methodological systems to address the industry pain point of 'lack of standards' for surface defects in optical manufacturing, promoting the localization process of inspection equipment. Concurrently, it offers systematic solutions for researchers and engineers in the field of optical imaging and defect detection, accelerating the innovative application of machine vision technology in ultra-precision manufacturing and propelling China's optical inspection technology into the ranks of international leaders.