Training Dataset supporting the publication: Single Shot Line-of-Sight Atmospheric Turbulence Profiling with STORM for Laser Satellite Communications

This is a training dataset for Speckle-based Turbulence Observation and Reconstruction via Machine Learning (STORM), a technique which aims at estimating atmospheric turbulence profiles from the measurement of a single star or laser beam speckle pattern on the ground. Used as to train the first STORM model through surrogate learning, this dataset was generated using a MATLAB wave propagation simulator to generate single star speckle patterns with the parameters described in the associated paper.It contains 50 speckles per 5064 turbulence profiles, thus a total of 253200 sample speckle files (sample_{profile#}-{speckle#}.h5) and 5064 tag files (sample_{profile#}_tag.h5).The github codes for the machine learning model, speckcn2, are documented on the following site:https://males-project.github.io/SpeckleCn2Profiler/.<br><br>

本数据集为基于散斑的机器学习湍流观测与重建（STORM）技术的训练数据集，该技术旨在通过地面单星或激光束散斑图样的测量，反演大气湍流廓线。本数据集用于通过代理学习训练首个STORM模型，其生成方式为借助MATLAB波传播模拟器，按照关联论文中所述参数生成单星散斑图样。每5064组湍流廓线对应50幅散斑图样，因此数据集总计包含253200个散斑样本文件（命名格式为sample_{profile#}-{speckle#}.h5）以及5064个标签文件（命名格式为sample_{profile#}_tag.h5）。本机器学习模型speckcn2的代码文档已发布于以下站点：https://males-project.github.io/SpeckleCn2Profiler/。