Research Of Optical Coherence Tomography In Strong Scattering Media Based On Reflection Matrix Method

Light propagating in a strongly scattering medium will be strongly scattered,causing severe distortion of the otherwise ordered wavefront phase.These strongly scattered lights produce randomly distributed constructive and destructive interference,eventually forming speckle.This prevents structural imaging and energy focusing of light inside biological tissue.In this paper,an optical coherence tomography system based on the reflection matrix method is used to research the imaging technology of light in strongly scattering medium,the phase extraction method of the reflection matrix is discussed,and the detected reflection matrix is optimized and regularized.The method of Kirchhoff regularization correction and the method based on the long short-term memory effect classifier are proposed to scientifically select the number of main singular values for image recovery,the imaging of a specific target light spot on the scattering medium is finally realized.The details are as follows:(1)Design the optical path diagram,build the complete experimental optical path.The relevant performance parameters of the optical devices used to modulate and load the phase in the optical path:the spatial light modulator SLM and the liquid crystal variable delay LCC were studied in detail,and the fitted grayscale-phase lookup table of spatial light modulator is obtained through pre-experiments.(2)In terms of phase extraction,the phase information of all complex amplitude elements used to form the reflection matrix is measured using the three-step phase shift method and the four-step phase shift method respectively.The optical reflection matrix of the strongly scattering medium is constructed and reorganized,and the optical reflection matrix that can be used for singular value decomposition is preliminarily obtained.By comparing the advantages and disadvantages of the two full-field phase-shift interferometry methods in terms of time complexity and image restoration quality,different detection methods are selected to measure the phase according to different situations.(3)The preliminarily extracted reflection matrix is normalized and filtered.The probability density distribution of the normalized matrix singular values should satisfy the standard quarter-circle distribution in statistical optics.The Kirchhoff regularization parameterλis introduced to scientifically select the number of eigenstates that can contribute to image reconstruction,avoiding the inaccuracy of the main singular value required for reconstructing image information based on empirical values.A resolution matrix_AT is introduced to test the weighting of the inversion results to the real model.Deep learning correlation training is carried out by long short-term memory effect classifier,and the optimal number of principal singular values is determined by combining the correlation between correlation curves and image pairs.