Generation Of Partially Developed Speckle And Investigation Of Imaging Through Scattering Medium

Since the invention of the laser in the early 1960s,people have begun to study the speckle phenomenon in optics.It has been found that when the laser is irradiated on a smooth wall or paper,a speckle image with a granular distribution is observed.Walls and paper are smooth on a macro scale,but when measured in the visible light band,the surfaces of these objects are completely rough.The reflecting surface is composed of countless tiny reflective primitives in different directions.When a beam with good coherence is irradiated on these rough surfaces,it is equivalent to random phase and direction modulation of the incident light field,and random speckle field occurs in the outgoing light field due to interference.According to the central limit theorem,it can be deduced that the amplitude of random speckle field statistically obeys the Rayleigh statistical distribution,and the intensity obeys a negative exponential distribution.In recent years,non-Rayleigh statistically distributed speckle fields have gradually attracted people's attention,which have played an important role in dynamic speckle illumination microscopy imaging,optical super-resolution imaging,and high-order correlation imaging based on pseudo-thermal light sources.In addition to the research on the physical properties of speckle fields,people have also conducted in-depth studies on the process of laser light transmitting through scattering media,such as focusing of laser light through scattering media and imaging through scattering media.In this paper,the following two contents are studied.The first is to use digital micromirror device to generate the non-Rayleigh statistical distributed speckle field.The second is to study the application of deep learning in the field of imaging through scattering media.The first main work of this paper is to use digital micromirror device to realize the generation of non-Rayleigh statistically distributed speckle fields.The related statistical properties of the non-Rayleigh distributed speckle field are studied and analyzed.Compared with liquid crystal spatial light modulators,the most prominent advantage of digital micromirror devices is the refresh rate.The refresh rate of liquid crystal spatial light modulation devices is generally between tens and hundreds of hertz,while the refresh rate of DMD can reach tens of thousands of hertz,and the refresh rate of the DMD used in our experiment can reach up to 22.7kHz.Using DMD to realize the generation of non-Rayleigh statistical distributed speckle field can greatly improve the efficiency of related applications based on non-Rayleigh statistical distributed speckle field,and expand its application in more fields.In this paper,a superpixel based method is used to achieve the modulation of complex light field by using DMD through an off-axis 4f system and spatial filtering.This method has a certain limitation on the spectral range of the target light field.Frequency components outside the system bandwidth cannot pass through the low-pass filter window at the middle spectral plane of the 4f system,which greatly reduces the modulation accuracy of the method.Therefore,in the process of generating the non-Rayleigh statistical distribution speckle field,in addition to considering the statistical distribution of the target speckle field,the spectral range of the target speckle field also needs to be considered.The second major work of this paper is to apply deep learning to the field of imaging through scattering media.The technology of imaging through scattering media plays a very important role,especially in the field of biomedical imaging.There are many traditional methods for imaging through a scattering medium,such as imaging through a scattering medium based on wavefront correction technology and imaging through a scattering medium based on the angular memory effect.In order to explore the new imaging through scattering media technology,this thesis conducts related research on the application of deep learning in the field of imaging through scattering media.The pictures in the handwritten digital recognition library are loaded on a spatial light modulator as the object,and the expanded and collimated laser beam is irradiated on the object.The transmitted beam is collected by the camera after passing through the scattering medium(ground glass)and the imaging system.The camera collects the speckle image of the object after passing through the scattering medium.It is impossible to directly obtain any useful information of the object from the speckle image.Using a kind of fully convolutional neural network(U-NET),through the training of a large amount of data,finally realized imaging through scattering media based on deep learning.