Complex Weight Matrix Based Optical Imaging And Light Control Through Complex Media

Deep imaging and light field control in complex media such as biological tissues have been hot topics in the field of optical research.However,light propagation in complex media is severely scrambled by multiple scattering,and it is challenging to directly image and manipulate the optical field through complex media.Imaging and focusing through complex media is achieved by modulating amplitude or phase of incident light using wavefront shaping.Machine learning has also been applied to image reconstruction and light field control in complex media.In this paper,we achieved image reconstruction and optical field control through complex media by constructing the mapping relationship between incident and pseudo-emitting light fields of complex media with complex weight matrix based on wavefront shaping and machine learning.First,we designed and built an experimental system for light transmission in complex media.We completed spatial light modulator phase calibration,signal synchronization of input and output,and system verification experiments.We verified the reliability of the optical experimental system by input image reconstruction of multimode fibers such as handwritten digital images with U-net and focusing through multimode fibers using genetic algorithms.Secondly,we combine the linear regression algorithm in machine learning with the optical transmission matrix theory to propose a complex weight matrix model.It is successfully implemented to reconstruct input image such as handwritten digital image and complex natural image from output speckle of multimode fiber by using tens of thousands of input images and output speckles to train the complex weight matrix.Finally,we performed emitted optical field control of multimode fibers utilizing the same approach of complex weight matrix model.Focusing at arbitrary positions,fast focus scanning,and image projection with arbitrary shapes were demonstrated at the end of the multimode fiber using the complex weight matrix obtained.The complex weight matrix model with multi-task learning capability can be used for both image reconstruction and optical field control.This work provides a new approach for imaging and optical field control utilizing scattered light,extending the application of wavefront shaping and machine learning in the fields of scattered imaging and optical transmission control.