Research On Obtaining Method Of Surface Height Information Of Glass Substrate

Large-size and ultra-thin glass substrates are one of the most important basic materials for the strategic emerging industry-the electronic information display industry.Excessive height changes occur during the production process,leading to defects in key components of information display devices,which directly affect the quality of the displayed graphics.It can even cause the production of defective products and waste products,so the research on the method of obtaining the height information of the glass surface is very important.At present,non-contact measurement is mainly used.Among them,spectral confocal technology has attracted more attention because of its rapidity and high accuracy in measurement.This paper proposes the acquisition of glass substrate surface height information based on the principle of spectral confocal measurement.Through theoretical analysis and software simulation,a set of spectral confocal measurement system is designed.At the same time,CCD camera is used to acquire spectral images to replace the spectrometer as the image acquisition of the measurement system.Lay a foundation for the follow-up study of the spectral confocal system.This article mainly studies the system through the following points:(1)The theory of the spectral confocal system is studied in detail,and the main factors affecting the performance of the system are analyzed.Analyze each component in the optical system,select reasonable hardware,and set optical design indicators.For the key component-the dispersion objective lens,the dispersion size is 500 μm and the relationship between the wavelength and the axial height is obtained by fitting.(2)In order to facilitate system integration and reduce system costs,a color CCD camera is used instead of a traditional spectrometer as a spectral image acquisition device.Aiming at the problem that the relationship between the RGB value and the wavelength of the central pixel of the image spectrum is difficult to convert,a color conversion model is proposed.The RGB value is converted into an XYZ color model by formula,and then through the corresponding relationship between chromaticity coordinates and wavelength in the XYZ color model,it is proposed to use BP neural network for data processing to find a mathematical model with higher measurement accuracy.(3)The designed optical path is used to build a system for obtaining information on the surface height of the glass substrate.First,it is verified that the colors of the corresponding spectral images at different positions are different.Then,according to the existing data,the BP neural network prediction is performed on the two,and the fitting results show that the BP neural network has high fitting accuracy.At the same time,a filter of a specific wavelength is selected to verify the mathematical model,and the average absolute error of the predicted wavelength is not more than 0.6 nm.And put the predicted wavelength into the fitting formula of wavelength and axial height,after eliminating the large error,obtained that the absolute error of the system measurement height position does not exceed 1.4 μm,and the range is 500 μm,which meets the industry’s micron-level measurement accuracy requirements.