基于小波包与支撑矢量机的天体光谱自动分类方法

针对天体光谱,通过小波包分解对光谱特征进行提取,并将该特征与支撑矢量机相结合,从而得到一种对活动天体与非活动天体实现自动分类的新方法.该方法未用到谱线信息,避免了谱线提取的复杂过程,使得在低信噪比,红移未知的情况下,依然能够对活动天体与非活动天体进行有效的分类识别.通过实验表明,该方法比其他现有方法有更高的识别率,可达到96.64%,并具有相当好的鲁棒性.

Automatic Classification Method for Astronomical Spectra Based on Wavelet Packet and SVM

The spectra features can be extracted by decomposition of wavelet packet for the celestial spectra.Combining the features with the Support Vector Machines(SVM),we obtained the effective classification methodology for non-active objects and active objects.Due to the information of spectral line was unemployed,the proposed method avoids the complexity of the extraction of spectral line and can effectively recognize the spectra of active objects when red-shift values is unknown and SNR is low.Experimental results show that the proposed method is more effective then other methods,the identification rate can be up to 96.64%,and the method has good robustness.