Phase Aberration Compensation Based On Deep Learning And Matrix Recovery

Digital holographic microscopy is a quantitative phase measurement technology based on optical interference and numerical diffraction computation,with its advantage of non-contact measurement,high reconstruction speed,high measurement accuracy,and so on,widely used in cell investigation,surface morphology,fluid analysis,and other fields.In digital holographic microscopy,phase aberration caused by off-axis structure and the micro objective lens is a key factor for its quantitative phase measurement accuracy,thus phase aberration compensation is an important and lasting topic in digital holographic microscopy research field.This dissertation firstly reviews the development history and current status of digital holographic microscopy,then introduces wave-front recording and numerical reconstruction principle,production of phase aberration,and phase aberration compensation method.Be concerned with phase aberration compensation problem in digital holographic microscopy,we proposed a phase aberration compensation method based on background segmentation and matrix recovery.Meanwhile,by implementing of background segmentation based on deep learning,we achieved automatic phase aberration compensation.The core contents and innovations of this dissertation can be summarized in the following two aspects:(1)Explored the low-rank property of phase aberration distribution,we proposed a phase compensation method based on background segmentation and matrix recovery,by segmenting the background area of complex amplitude distribution of holograms,the proposed method avoids influence caused by phase information of the object in the matrix recovery process,solved the phase aberration compensation problem in the scene of measuring big object to a certain extent,the effectiveness of the proposed method is proved by both numerical simulative and optical experiment.(2)Furthermore,we designed a neural network,achieved automatic processing of background segmentation.Besides,we proposed a generating method of simulated hologram training dataset by frequency domain fusion method,holograms in the training dataset have similar background noise distribution with real holograms,achieved holograms dataset augmentation.Combined the designed neural network with the phase aberration compensation method which is proposed in(1),we realized automatic phase aberration compensation for the digital holographic microscopy imaging system.