Detection Technology Of Quartz Glass Internal Defects Based On In-line Digital Compressive Holography

Quartz glass has a widespread application in numerous fields—national defense,optoelectronic devices,chemical industry and etc.—due to its excellent heat resistance,spectral transmittance and insulation property.Structural defects including bubbles,granules and strips are produced when glass is forming.These defects could seriously affect the performance of glass.The detection of internal defects without damaging the glass is of great importance.In-line digital holography is wildly used in object measurement.While it might on one hand be outstanding in non-contact,simple structure and high resolution,on the other hand its zero-order terms,real and conjugated images are aliasing together,also the information of real image is not legible.Traditional methods such as phase shift and off-axis holography can eliminate zero-order terms and conjugated images.However,phase shift sacrifices the speed of the imaging system to capture multiple holograms,and off-axis holography sacrifices the spatial bandwidth product.How to eliminate zero-order terms and conjugate images on the foundation of only collecting a single Gabor hologram,to detect internal defects of quartz glass with high resolution is the focus of this paper.Subject "Detection Technology of Quartz Glass Internal Defects Based on In-line Digital Compressive Holography" is a study on the in-line digital holographic imaging algorithm based on compressed sensing theory.Aiming at solving the boundary noise problem in the block compressive holography algorithm,an uneven block model which combines with zero-padding and redundant pixel-padding is proposed.This model not only ensures the speed advantage of the block-wise parallel algorithm,but also achieves the high-resolution reconstruction of hologram of bubble sample.Firstly,on the basis of analyzing the principle of compressed sensing and digital holography technology,an in-line compressed holographic imaging model is established,the principle of compressed holographic imaging is discussed and its key factors are analyzed.Aiming at the positioning and measurement of glass bubble samples,the hologram denoising algorithm and threshold binarization algorithm are studied.Then,the superiority of Two-Step Iterative Shrinkage Thresholding algorithm and total variation regularization operator in the field of image reconstruction is proved via the compressed sensing reconstruction experiment of one-dimensional signal and two-dimensional image.On the basis of equal-divided sub-hologram,an uneven block compressive holography model is proposed,and the redundant information protects the boundary of the sub-hologram.The model ensures the compressed holographic imaging quality and highlights the speed advantage of the block-wise parallel algorithm.In the resolution target simulation and polystyrene particle experiment,a variety of in-line holographic imaging algorithms are compared,which verifies the superiority of the proposed algorithm model.Lastly,the inside bubble detection device for the quartz glass is built,and the lateral resolution of the system and the standard diameter of the bubble are determined by applying the precision scale and the bubble focusing hologram.The influence of different padding modes on the reconstruction results of bubble samples with the uneven block compressive holography algorithm is analyzed.After the threshold binarization of the reconstruction results,the diameter,area,and position of bubbles can be obtained by the connected domain.According to the proposed algorithm,a set of software interactive interface is designed,which integrates the steps of removing direct current,autofocusing,compressed holographic reconstruction,threshold binarization and other steps in the bubble detection process,and users can select the defocused bubble hologram that needs to be processed according to their demands,and finally obtain its reconstruction image and bubble information directly from the system,and the interface is well-ordered and user-friendly.