一种基于小波变换的晚型星自动识别方法  

A Wavelet-Transform-Based Method for the Automatic Detection ofLate-Type Stars

作　　者：刘中田[1] 赵瑞珍[2] 赵永恒[3] 吴福朝[1] 

机构地区：[1]中国科学院自动化研究所模式识别国家重点实验室,北京100080 [2]北京交通大学计算机与信息技术学院,北京100044 [3]中国科学院国家天文台,北京100012

出　　处：《光谱学与光谱分析》2005年第7期1158-1161,共4页Spectroscopy and Spectral Analysis

基　　金："863"计划(2003AA133060);国家自然科学基金(60202013)资助项目

摘　　要：我国正在执行的大型巡天计划(LAMOST计划),急需晚型星的自动识别系统。然而,在文献中尚未见到关于对晚型星有效自动处理方法的报道。文章旨在探索研究对晚型星自动处理的有效方法。在文章中,晚型星主要是指具有较强分子吸收带的恒星,主要包括富氧的M,L和T型星以及富碳的C星。通过研究我们发现:对晚型星光谱进行5层小波变换后,第5层小波系数的频谱具有单峰分布的特性,其能量主要集中在以单峰为中心的特征区间内;对于其他类型的光谱,相应的频谱具有多峰分布的特点,其频谱能量比较分散,它们在晚型星特征区间内的频谱能量非常有限。基于上述发现,给出了一种基于小波变换的晚型星自动识别方法。实验表明该方法是切实可行的并且具有相当好的鲁棒性。The LAMOST project, the world largest sky survey project, urgently needs an automatic late-type stars detection system. However, to our knowledge, no effective methods for automatic late-type stars detection have been reported in the literature up to now. The present study work is intended to explore possible ways to deal with this issue. Here, by “late-type stars” we mean those stars with strong molecule absorption bands, including oxygen-rich M, L and T type stars and carbon-rich C stars. Based on experimental results, the authors find that after a wavelet transform with 5 scales on the late-type stars spectra, their frequency spectrum of the transformed coefficient on the 5^th scale consistently manifests a unimodal distribution, and the energy of frequency spectrum is largely concentrated on a small neighborhood centered around the unique peak. However, for the spectra of other celestial bodies, the corresponding frequency spectrum is of multimodal and the energy of frequency spectrum is dispersible. Based on such a finding, the authors presented a wavelet-transform-based automatic late-type stars detection method. The proposed method is shown by extensive experiments to be practical and of good robustness.

关 键 词：小波变换 晚型星 天体光谱 分子吸收带 

分 类 号：TN911.7[电子电信—通信与信息系统]