小行星的物质组成提供了关于小行星的起源和演化的重要信息，以及对太阳系内层行星形成理论的约束。由于地面观测仪器的限制，目前缺乏暗弱的小行星的物质组成信息。不久的将来，中国空间站望远镜（CSST）将获得亮度超过25 mag和23 mag的小行星的多色测光数据与无缝光谱数据。利用所获得的光谱数据和多色测光数据，可以研究小行星的表面物质组成。小行星的表面物质组成不同，其反射光谱的特征也不同。因此，按照反射光谱的形状及吸收特征将小行星分成不同的类型。未来CSST将为小行星的物质组成研究带来海量观测数据，如何应用CSST的光谱数据开展小行星物质组成分析是本论文主要解决的问题。为此，借助机器学习技术建立了一整套适用于未来CSST小行星无缝光谱数据的人工神经网络（Artificial neural networks, ANN）分析工具。根据CSST无缝光谱模块的设计指标，使用SMASS II的光谱数据和Bus-Binzel分类系统的标签构建了由5个单独的人工神经网络组成的集成分类工具。对测试集中数据的分类的准确率达到92 %以上，对S3OS2（Small Solar System Objects Spectroscopic Survey）数据集的分类测试的准确率超过87%。作为首次应用，使用该人工神经网络分类工具分析了2006年和2007年在中国兴隆观测站（天文台代码327）使用2.16米望远镜观测的42颗小行星的64条光谱，给出了64条光谱的类型并与以往的研究进行比较。结果表明，小行星（469）的表面物质成分是均匀的（除了零相位附近存在1条C类光谱），并首次给出了小行星（1303）的光谱类型（D类）。此外，发现64条光谱中有23条光谱在0.7μm附近存在吸收特征，这些小行星的表面可能存在含水矿物。从测试结果与实际应用中的分析结果来看，这一人工神经网络分类工具可以应用于分析未来CSST的小行星无缝光谱，能够满足小行星物质组成的研究需求。
其他摘要
The composition of asteroids gives important information to understand their origin and evolution, as well as restrictions on the formation of planets in the inner solar system. However, we lack composition information for faint asteroids due to observation limits with ground-based instruments. In the near future, the survey observaions with Chinese Space Survey telescope (CSST) will provide multiple colors and spectroscopic data for asteroids of apparent magnitude brighter than 25 mag and 23 mag, respectively. Using the obtained spectroscopic data and multiple colors , the surface material composition of asteroids can be studied.The surface composition of asteroids is different, and the features of their spectra are also different. Therefore, asteroids are classified into different types according to the shape of the reflection spectrum and absorption features. In the future, CSST will bring a large amount of observational data for the study of asteroid material composition. How to use the spectral data of CSST to analyze asteroid material composition is the main problem to be solved in this work.Therefore, we developed a classification tool applied a machine learning technique tools suitable for future CSST asteroid slitless spectral data. According to the design of CSST slitless spectrum module, an integrated classification tool consisting of 5 separate artificial neural networks is constructed using the spectral data of SMASS II and the labels of Bus-Binzel taxonomy system. The accuracy of analyzing the data of the test dataset is more than 92%, and the accuracy of the classification test on the S3OS2 (Small Solar System Objects Spectroscopic Survey) dataset was more than 87%. As the first application of our ANNs (Artificial neural networks) tool, 64 spectra of 42 asteroids obtained in 2006 and 2007 with the 2.16-m telescope in the Xinglong station (Observatory Code 327) of National Astronomical Observatory of China are analyzed, giving the types of 64 spectra and comparing those results with previous studies. We find that the surface material composition of asteroid (469) is uniform (except for the presence of a C-type spectrum near zero phase), and for the first time give the spectral type (D-type) of asteroid (1303). Additionally, we found that 23 of the 64 spectra have absorption feature near 0.7μm, and there maybe aqueous minerals on these asteroids’ surface.Considering the test results and the results in application to the real observation data, the ANNs classification tool will be used to analyze the futrue CSST spectra data of asteroids, which could bring breakout results in asteroids’ origin and evolution.
修改评论