Open AccessIdentification of Interference Sources in Spiral Galaxy Images Based on Convolutional Neural Network
Author(s) -
Shiqing Wu,
Hu Tao
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1213/4/042006
Subject(s) - convolutional neural network , normalization (sociology) , galaxy , computer science , artificial intelligence , spiral galaxy , stars , pattern recognition (psychology) , astrophysics , artificial neural network , physics , algorithm , sociology , anthropology
In the spiral galaxy image, the luminosity of the precursor star due to its close proximity to the spiral galaxy reaches an approximate order of magnitude, which makes the precursor stars become an important interference source in the galaxy image and makes the computer difficult to distinguish them. Therefore, it is of great significance to remove the precursor stars from the galaxy image. In this paper, a modified convolution neural network (CNN) is used to classify spiral galaxies and stars in order to find pre-star interference. Convolutional neural network (CNN) is a popular image classification technology, which is included in depth learning. It solves the problem of a large amount of data and serious non-linearity on the basis of neural network. Based on CNN, this paper studies and improves the selection of network level, the optimization algorithm, batch normalization and so on. The two types of astronomical images of the target are trained and predicted, and the classification results are very good. At present, the error rate is around 3%.