Research On Super-resolution Reconstruction Technology Of Dark-field Image Of Optical Element Damage

In order to improve the initial damage detection and tracking capabilities of large-aperture optical components in high-power laser devices,in view of the problem of low texture information and weak key signals in dark-field images of optical component damage,a super-resolution reconstruction of dark-field images based on model integration is proposed.The method is of great significance for the early detection of optical component damage.The main research contents of this paper are as follows:(1)A systematic review of image super-resolution reconstruction methods,and analysis of the principles and characteristics of image super-resolution reconstruction techniques from traditional interpolation methods to current deep learning methods;(2)A method for super-resolution reconstruction of dark-field images based on model integration is proposed.This method adopts the model integration method of image weighted fusion and integrates two deep learning algorithms: the overall attention model and the divide-and-conquer model.The overall attention mechanism includes The layer attention module and the channel space attention module,the layer attention module focuses on the relationship between the levels,and the channel space attention module focuses on the confidence of each channel position and the relationship between the channels.Then tap the channel and the correlation between channels.The divide-and-conquer model consists of an attention block and a stacked hourglass model,and learns multi-scale features through bottom-up and top-down reasoning;(3)Build an image acquisition system for optical component damage.Using100 mm fixed focal length lens and zoom lens self-made "down-sampling" and "real image" two high-low resolution dark field image data sets S＿DATA;(4)The Pytorch deep learning framework is used to integrate the overall attention model and the divide-and-conquer model based on image weighted fusion,and achieve 2～4 times super-resolution reconstruction results;(5)Using subjective evaluation,peak signal-to-noise ratio PSNR,structural similarity SSIM and other indicators,the reconstruction results of dark-field images of optical element damage were evaluated.The experimental results prove the effectiveness of this method.