基于卷积神经网络的大气中光路气流扰动实验研究  

Atmospheric Optical Path Airflow Disturbance Analysis Method Based on Convolutional Neural Network

作　　者：刘一琛 吴侃[1] 邱高峰 陈建平[1] Liu Yichen;Wu Kan;Qiu Gaofeng;Chen Jianping(State Key Laboratory of Advanced Optical Communication Systems and Networks,Shanghai Jiao Tong University,Shanghai 2002A0,China)

机构地区：[1]上海交通大学区域光纤通信网与新型光通信系统国家重点实验室

出　　处：《光学学报》2019年第8期8-18,共11页Acta Optica Sinica

基　　金：国家自然科学基金(61505105,61875122)

摘　　要：提出了一种基于激光光斑畸变和卷积神经网络(CNN)的光路气流扰动研究方案。利用CNN对激光光束在空间传播中受到气流扰动后的光斑畸变进行学习,得到光束传播路径上的气流扰动情况。实验表明,训练得到的评估参数与由风速仪测得的光路中的气流扰动(风速)具有强相关性。本方案提供了一种短距离、快速、低成本的气流扰动分析手段。A method to investigate optical path turbulence based on laser spot distortion and a convolutional neural network(CNN)is proposed.Utilizing the CNN,we evaluated the spot distortion of laser beams resulting from airflow disturbance in space propagation.As a result,details of turbulence on the beam propagation path can be obtained.Experimental results demonstrate a high correlation between the evaluation parameter and the turbulent intensity(wind speed)measured by an anemoscope.The proposed method provides a turbulence analysis with short distance,high speed,and low cost.

关 键 词：大气光学 空间光学 气流扰动 卷积神经网络 深度学习 

分 类 号：P412.16[天文地球—大气科学及气象学]