Disorderly Dispersion Based On Computational Reconstruction Research And Improvement Of Miniature Spectrometer

As a compact and highly accurate spectroscopic analysis device,miniature spectrometers have been widely used in modern material analysis fields for measuring and identifying the optical properties of materials.With the growing demand for real-time detection and relying on micro-nano manufacturing technology,efficient computing algorithms,and the development of artificial neural networks,it has become increasingly important to develop low-cost and widely applicable miniature spectrometers.This paper conducts relevant research on the design and optimization of disorderly dispersed miniature spectrometers based on computational reconstruction,with the main content including:1、Design and experimental research of the computationally reconstructed disorderly dispersed miniature spectrometer: Based on the spectroscopic principle of disordered frosted glass paper,a miniature spectrometer suitable for the 400-800 nm wavelength range was designed.While ensuring the performance of the miniature spectrometer,low-cost disordered frosted glass paper was used to complete the theoretical design of the miniature spectrometer,and experimental verification was carried out.2 、 Algorithm optimization of computationally reconstructed disorderly dispersed miniature spectrometers: A detailed analysis of the ill-conditioned equation system in the computational reconstruction of miniature spectrometers caused by system noise is presented.Singular value decomposition algorithms and Tikhonov regularization algorithms are used for theoretical analysis and experimental verification of spectral reconstruction.The optimization of the above two algorithms is achieved using the idea of locally weighted regression smoothing,resulting in a peak shift of 3.95 nm and the full width at half maxima(FWHM)difference of 2.66 nm for the reconstructed spectral curve.3、Theoretical research and experimental demonstration of miniature spectrometers based on artificial neural networks: Analyzing the influence of internal and external noise on the computationally reconstructed miniature spectrometer system,artificial neural networks are used to improve the instability and fluctuation of the reconstructed spectral results of traditional computational algorithms.By utilizing different artificial neural network training algorithms to train the network as a substitute for the matrix equation of the computational reconstruction miniature spectrometer system,the accuracy of the reconstructed spectrum is improved,resulting in a peak shift of 1.58 nm and the full width at half maxima(FWHM)difference of 0.83 nm at the optimal point of the reconstructed spectral curve.