Investigation On Light Field Imaging Passing Through Hollow Fiber Based On Deep Learning

In recent years,deep learning technology has been developed rapidly and is extending to different fields.Traditional image recognition or speckle reconstruction techniques such as transmission matrix technology and wavefront shaping technology have certain complexity in operation and require a stable optical system to obtain good results.The application of deep learning technology to optical imaging can accurately extract image features and restore images,and the experimental operation is simple.The main work of this thesis is listed as follows:1.Analyzed what is deep learning,the source of deep learning,introduced the basic ideas of deep learning,and described the commonly used deep learning network architecture.The network structure of convolutional neural network and the working mechanism of each structure and its role in the network are explained in detail.2.The basic structure of the optical fiber is introduced.The basic principles of light transmission in step fiber and gradient fiber are introduced respectively and explained from the perspective of geometric analysis.Then I made a report on the development of scattering medium imaging and introduced several methods of scattering medium imaging in detail.3.Using deep learning technology,the effects of hollow core optical fibers with different lengths and different apertures on information transmission were explored.It was found in experiments that the longer the length of air-core fiber under the same aperture,the better the image restoration effect using U-shaped convolutional neural network.And the hollow fiber with the larger aperture at the same length,the higher the similarity between the restored image and the original image.The transmission and network recovery of the image information loaded by the amplitude-type spatial light modulator is significantly better than the image information loaded by the phase-type spatial light modulator for transmission,and the former is easier and simpler in network training.