Model for restoring obstructed beam transmission in atmospheric turbulence based on BP neural network

Abstract

Atmospheric turbulence and obstacles can distort rays during transmission, resulting in significant wavefront distortion and loss of optical field information. This paper employs the phase screen method to simulate the transmission characteristics of a Gaussian plane wave in turbulent conditions, establishing an obstacle grid at the receiver to represent beam obstruction. A dataset of unobstructed transmissions is used to train a Backpropagation Neural Network, constructing neurons and connection weights. By scanning optical field data systematically, the model compensates for the obstructed portions of the optical field distribution. The results are compared to unobstructed transmissions, focusing on image similarity, and demonstrate the entire process from compensation to distortion correction. Simulation results indicate that the Backpropagation Neural Network effectively compensates for optical field information loss, showcasing strong performance within a certain time scale.