Huang Shuan-Yu, Karmakar Riya, Chen Yu-Yang, Hung Wei-Chin, Mukundan Arvind, Wang Hsiang-Chen
Department of Optometry, Central Taiwan University of Science and Technology, Taichung 40601, Taiwan.
Department of Mechanical Engineering, National Chung Cheng University, Chiayi 62102, Taiwan.
Sensors (Basel). 2024 Aug 6;24(16):5094. doi: 10.3390/s24165094.
This study introduces a novel method for detecting and measuring transparent glass sheets using hyperspectral imaging (HSI). The main goal of this study is to create a conversion technique that can accurately display spectral information from collected images, particularly in the visible light spectrum (VIS) and near-infrared (NIR) areas. This technique enables the capture of relevant spectral data when used with images provided by industrial cameras. The next step in this investigation is using principal component analysis to examine the obtained hyperspectral images derived from different treated glass samples. This analytical procedure standardizes the magnitude of light wavelengths that are inherent in the HSI images. The simulated spectral profiles are obtained using the generalized inverse matrix technique on the normalized HSI images. These profiles are then matched with spectroscopic data obtained from microscopic imaging, resulting in the observation of distinct dispersion patterns. The novel use of images coloring methods effectively displays the thickness of the glass processing sheet in a visually noticeable way. Based on empirical research, changes in the thickness of the glass coating in the NIR-HSI range cause significant changes in the transmission of infrared light at different wavelengths within the NIR spectrum. This phenomenon serves as the foundation for the study of film thickness. The root mean square error inside the NIR area is impressively low, calculated to be just 0.02. This highlights the high level of accuracy achieved by the technique stated above. Potential areas of investigation that arise from this study are incorporating the proposed approach into the design of a real-time, wide-scale automated optical inspection system.
本研究介绍了一种利用高光谱成像（HSI）检测和测量透明玻璃板的新方法。本研究的主要目标是创建一种转换技术，能够准确显示从采集图像中获取的光谱信息，特别是在可见光光谱（VIS）和近红外（NIR）区域。当与工业相机提供的图像配合使用时，该技术能够捕获相关光谱数据。本研究的下一步是使用主成分分析来检查从不同处理的玻璃样品中获得的高光谱图像。这一分析过程使HSI图像中固有的光波长大小标准化。使用广义逆矩阵技术对归一化的HSI图像进行处理，从而获得模拟光谱轮廓。然后将这些轮廓与从微观成像获得的光谱数据进行匹配，从而观察到明显的色散模式。图像着色方法的新颖应用有效地以视觉上显著的方式显示了玻璃加工片的厚度。基于实证研究，近红外-高光谱成像范围内玻璃涂层厚度的变化会导致近红外光谱内不同波长的红外光透射率发生显著变化。这一现象是薄膜厚度研究的基础。近红外区域内的均方根误差低得惊人，计算结果仅为0.02。这突出了上述技术所达到的高精度水平。本研究引发的潜在研究领域包括将所提出的方法纳入实时、大规模自动光学检测系统的设计中。
