A Neural Network Method for Inversion of Turbulence Strength

Abstract

Accurate inversion of atmospheric turbulence strength is a challenging problem in modern turbulence research due to its practical significance. Inspired by transfer learning, we propose a new neural network method consisting of convolution and pooling modules for the atmospheric turbulence strength inversion problem. Its input is the intensity image of the beam and its output is the refractive index structure constant characterizing the atmospheric turbulence strength. We evaluate the inversion performance of the neural network at different beams. Meanwhile, to enhance the generalisation of the network, we mix data sets from different turbulence environments to construct new data sets. Additionally, the inverted atmospheric turbulence strength is used as a priori information to help identify turbulent targets. Experimental results demonstrate the effectiveness of our proposed method.