RFI clustering study of Yunnan 40 meters radio telescope based on improved SpectralNet
To develop specific mitigation strategies for addressing the issue of severe radio frequency in-terference(RFI)contamination during pulsar observations,an image clustering model called CAE-Spectral-Net,based on autoencoders and spectral clustering,has been proposed.The autoencoder part of this model automatically extracts key features from images,avoiding the difficulties of manual operations.Spectral cluste-ring clusters the closest points in feature space,which helps to reveal the underlying structural information of the data.Experiments were performed on 2 000 pulsar time domain and frequency domain image data collected by the Yunnan Observatory's 40 meters radio telescope.The results show that after improving the structure of the autoencoder,the SpectralNet model significantly improves on the internal indicators of clustering compared to the original SpectralNet model and some traditional clustering algorithms.The clustering results preliminari-ly achieve the goal of RFI classification.
CAE-SpectralNet modelimage clusteringmorphological classification of RFI
万方数据
维普
