本项目拟采用全球发布的LAMOST巡天和GAIA、SDSS、Pan-STARRS等多数据集的测光数据，首先通过颜色判据筛选出表面有效温度一般大于10000K的蓝超巨星，将筛选出的高温星与LAMOST进行交叉获得研究光谱范围，运用深度学习方法，通过已知的蓝超巨星观测光谱和理论光谱进行训练，搜寻蓝超巨星，利用光谱中的谱线来证认蓝超巨星。学习获取LAMOST观测中分辨、低分辨率光纤光谱数据出现机率较低的稀有天体，结合恒星天体空间的距离位置和特征值间的相关性分析后，探求恒星光谱物理参数间的相关性，提出海量光谱数据准确且多层次的稀少特殊天体分类，科学推测银河系演化过程。研发出特殊恒星大气物理参数相关性自适应算法软件包，探索深度学习完成特殊恒星的大样本搜寻、分类和天体物理特性的研究工作，加深对恒星结构乃至整个宇宙演化的认知理解，提供恒星结构和演化研究领域新方法，拓展LAMOST巡天数据的科学产出。

This project is intended to adopt the photometry data of multi-data sets of sky surveys such as LAMOST, GAIA, SDSS and Pan-STARRS, which are published worldwide. Firstly, the blue supergiants with effective surface temperatures generally greater than 10000K ,which are screened through the color criterion. Then, the selected hot stars are crossed with LAMOST to obtain the research spectral range. Using the deep learning method to train and know the observation spectrum and the theoretical spectrum of the blue supergiants , After learning to acquire rare objects with low probability of occurrence of resolved and low-resolution optical fiber spectral data in LAMOST observation, combining with the correlation analysis between the distance position of stellar objects and the characteristic values. To explore the correlation among physical parameters, spectra of stars mass spectral data accurately and multi-level rare special object classification, scientific inference the galaxy evolution. Developing special star correlation between atmospheric physical parameters and adaptive algorithm software packages, the deep learning methods complete special large sample of the stars search features, classification, and astrophysics research, deepen the cognitive understanding of the structure and even the entire universe evolution of stars, provide a new method of stellar structure, and amplify the research in order to expand the theories of LAMOST scientific achievements .