Classifying Spectra Based on DLS and Rough Set
Bo Qiu, Zhanyi Hu, Yongheng Zhao · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2003
Until now, it is still difficult to identify different kinds of celestial bodies depending on their spectra, because it needs a lot of astronomers’ manual work of measuring, marking and identifying, which is generally very hard and time-consuming. And with the exploding spectral data from all kinds of telescopes, it is becoming more and more urgent to find a thoroughly automatic way to deal with such a kind of problem. In fact, when we change our viewpoint, we can find that it is a traditional problem in pattern recognition field when considering the whole process of dealing with spectral signals: filtering noises, extracting features, constructing classifiers, etc. The main purpose for automatic classification and recognition of spectra in LAMOST (Large Sky Area Multi-Object Fibre Spectroscopic Telescope) project is to identify a celestial body’s type only based on its spectrum. For this purpose, one of the key steps is to establish a good model to describe all kinds of spectra and thus it will be available to construct some excellent classifiers. In this paper, we present a novel describing language to represent spectra. And then, based on the language, we use some algorithms to extract classifying rules from raw spectra datasets and then construct classifiers to identify spectra by using rough set method. Compared with other methods, our technique is more similar to man’s thinking way, and to some extent, efficient.