Research On Focusing Through Scattering Media Based On Random Measurement Basis And Genetic Algorithms

After passing through a medium with non-uniform refractive index,light experiences random changes in phase and amplitude,resulting in irregular speckle patterns at the observation surface.This scattering phenomenon disrupts the information carried by incident light,leading to negative impacts on the imaging quality of optical systems and optical communication.To eliminate these negative effects,researchers have proposed numerous theories and methods.Wavefront shaping is a promising method that adjusts the phase or amplitude of incident light to eliminate or utilize the effects of scattering media,achieving focusing or imaging through scattering media.Wavefront shaping techniques can be roughly divided into three categories: optical phase conjugation,transmission matrix method,and feedback-based optimization method.This article focuses on the transmission matrix method and feedback-based optimization method and attempts to propose two new algorithms for focusing through scattering media.Firstly,we propose a real part measurement method for the transmission matrix based on the pseudo-inverse matrix principle to address the issue of the current transmission matrix measurement method requiring a reference light path and special measurement basis.This method can measure the real part of the transmission matrix and obtain the corresponding mask by binary operation.By modulating the wavefront with this mask,single-point and double-point focusing through scattering media can be achieved.This method named the real-valued transmission matrix method,which uses a very simple experimental setup and does not require an additional reference light path,while any data can be used as a measurement base.This greatly improves the practicality of the transmission matrix measurement method.Secondly,traditional genetic algorithms suffer from slow convergence and instability.In this thesis,we propose an improved genetic algorithm that combines the real-valued transmission matrix method based on random measurement bases,gradient descent,and genetic algorithms.Compared with the real-valued transmission matrix method based on random measurement bases,this algorithm has lower computational resource requirements and can achieve better focusing effects.Compared with traditional genetic algorithms,it has faster convergence and more stable results.Meanwhile,by designing parallel measurement and computation,the required optimization time is reduced,and the overall time consumption falls between that of the real-valued transmission matrix method and genetic algorithms.Both methods proposed in this article reduce the experimental conditions required for wavefront shaping and increase the possibility of practical applications.