Automatic Search And Study Of Planetary Nebulae And HII Region Based On Multi-source Data Intersection

Planetary nebulae and HⅡ region are a pair of special celestial bodies in the universe,which happen to be the origin and destination of some stars in their lifetimes.They are important probes for studying stellar evolution,interstellar medium and galaxy chemical formation history and evolution.Both of them are rare celestial bodies,limited by the number of samples,and related research is greatly restricted.It is very meaningful and necessary to expand new rare celestial body candidates and their data samples.The Large Sky Area Multi-Object Fiber Spectroscopic Telescope(LAMOST)has started the observation mode of low resolution and medium resolution alternately,and the new sky survey data contains huge research value.The bands covered by the LAMOST medium-resolution spectrum are very suitable for searching for celestial objects with Hα6563(?) and[OⅢ]5007(?) emission line features,providing an excellent data source for the search of planetary nebulae and HⅡregions.To this end,an effective algorithm should be designed to efficiently screen candidates for planetary nebulae and HⅡ regions from massive,complex,diverse,and unbalanced data.The physical properties and formation processes of planetary nebulae and HⅡ region are different,but their spectra are very similar.In the process of medium resolution spectral data mining,we regard them as the same positive sample for astronomical data mining.For the further distinction between planetary nebulae and HⅡ region,only spectral data cannot be achieved.It is necessary to combine multi-band data,such as WISE,Spitzer and other infrared telescope data,to distinguish them by photometry.In this thesis,we use the data of different bands from multiple astronomical databases(multi-source)to mining and further distinguish planetary nebula and HⅡ region on the LAMOST DR7 medium-resolution data to amplify new rare object candidates and their data samples.The main work and innovations of this thesis are as follows:(1)Planetary nebulae and HⅡ region star catalog information was collected to obtain sufficient coordinate information of target celestial bodies;make full use of multi-source astronomical databases,and cross-match the spectral samples of target celestial bodies as much as possible to construct sufficient positive sample data sets,which are in total in the LAMOST and SDSS databases interleaved to 966 spectral data of target celestial objects.For low-resolution spectra,the cubic spline interpolation method is used for data upscaling and a unified system to expand the dataset for medium-resolution search tasks.(2)A reasonable deep learning model was built and designed,and introduce self-calibration convolution to improve the feature extraction capability of the model;in pursuit of high recall rate performance while considering the balanced performance of other performance of the model,a series of parameter adjustments have been made to the model.A reasonable proportion of positive and negative training samples and ensemble learning are used to optimize the data imbalance problem.After model screening,low-quality spectral elimination,and manual review,a total of 2,649 spectral features matching HⅡ region and planetary nebulae were found.(3)Based on the different color characteristics of planetary nebulae and HⅡ region in the infrared band,the infrared images of the candidate targets were cross-matched.Through aperture photometry of the obtained infrared images,7 HⅡ region targets and 31 planetary nebulae targets were determined.