Recognition of Astronomical Strong Gravitational Lens System Based on Deep Learning

Dark matter is an important part of cosmological matter, and gravitational lensing is a fundamental way to study the distribution of dark matter. Therefore, the gravitational lensing system is of great significance to the study of astrophysics. However, since strong gravitational lensing is extremely rare, the number of strong lensing candidate galaxies within the galaxy is also very large. To find these rare objects, we need to find them from at least tens of millions of images. The features based on the shallow structure are difficult to perfectly process these broad sense image data. In order to improve the efficiency and accuracy of recognition, a deep learning-based astronomical method for astronomical strong gravitational lensing systems is proposed, which first requests the classification of image data using label files. Next, data preprocessing is used for augmentation of targets in images, which provides more samples and more features can be learnt. After classification and preprocessing, the neural network is trained with the data, so as to facilitate the detection of strong lenses. By actual testing, the experiments show that the results of our method is 4.88% higher than Kapteyn Resnet and 7.5% higher than Manchester-SVM in model evaluation index AUC.