Research On LAMOST M Dwarf Sub-Classification Based On Residual Distribution Measurement

With the great improvement of instrumentation and data collection ability, massive data is generated from some sky surveys, such as the LAMOST survey. How to process and analyze such big data sets is an important issue for astronomy. Based on the analysis and processing of the spectral data, we can determine the physical parameters of stellar atmosphere and the spatial distribution of spectrum.The classification of stellar spectra is an important job in data processing of astronomy, which is mainly used for searching celestial spectra with known types in massive data survey. Training a classifier by the given two and more categories of spectrum, gives a classification for new a spectrum. Template library can be treated as a trained classifier. We can get the classification results by template matching.This paper focused on LAMOST M dwarfs fine classification based on measurement of residual distribution. Residual distribution measurement is a measurement method used to measure the distance between two spectra. In the process of calculating the distance between two spectra, normalized processing should come first. Then calculate the residuals of the sampling points of corresponding wavelength, and eventually calculate the standard deviation of the residual distribution as the distance between the spectra. Before the classification of spectra, the first job is to process the spectrum with interpolation, normalized spectral and other processing operations. Calculate the distance between the spectra and measure each spectrum in the template library using distance measurement methods with residual distribution. We searched a template spectrum with a minimum distance, and then took the type of this template spectrum as the type of the spectrum to be measured. In this paper, the simulated spectra and the actual spectrum are used as the experimental data of classification. The simulated spectra are generated by the template spectra, using the M star of LAMOST DR2. The experimental results show that the spectra data can be classified more accurately with the measurement method of residual distribution than the use of other traditional classification methods. The effect of spectral classification is affected by signal to noise ratio, outliers, residual standardized coefficient and other factors. The larger the signal to noise ratio is, the better the spectral classifications will be.All in all, classifying the celestial spectra through the residual distribution measurement is feasible. The classification method and its process in this paper need to be further studied so that better classification effect could be achieved in wider use in the future.