基于联合模型的脉冲星候选体筛选方法研究  

Screening Method of Pulsar Candidate Based on Joint Model

作　　者：石纯高 袁浩然 赵一 王洪丰 SHI Chungao;YUAN Haoran;ZHAO Yi;WANG Hongfeng(School of Information and Control Engineering,Jilin Institute of Chemical Technology,Jilin Jilin 132022,China;School of Computer and Information,Dezhou University,Dezhou Shandong 253023,China)

机构地区：[1]吉林化工学院信息与控制工程学院,吉林吉林132022 [2]德州学院计算机与信息学院,山东德州253023

出　　处：《德州学院学报》2025年第2期7-12,18,共7页Journal of Dezhou University

摘　　要：作为一种快速旋转的中子星,脉冲星除了具有强磁场以及强引力的作用之外,其自身的特点是具有稳定的自转周期。现代射电望远镜数据量的增加,搜索到大量的候选样本,人工智能技术在脉冲星候选识别领域为解决海量数据的识别发挥着日益重要的作用。因此,基于深度学习对脉冲星候选者辨识精确度进行提升,也成为了一个重要的挑战。该方法设计了一个联合模型用以鉴别从五百米单口径射电望远镜(FAST)数据中收集的候选者,这一联合模型利用ConvNeXt网络和DenseNet网络,对候选者的二维数据和一维数据分别进行特征学习,然后使用全连接层对最终的分类结果进行判别。该模型对来自FAST的数据集测试中,召回率和精确率分别达到94.8%、95.7%,相比较于V-CCNN(垂直连接卷积神经网络)和H-CCNN(水平连接卷积神经网格)方法在召回率以及精确率上均有所提升。As a rapidly rotating neutron star,pulsars are characterized by their stable rotation periods in addition to having strong magnetic fields and gravitational effects.With the increase in data from modern radio telescopes,a large number of candidate samples have been discovered.To address the challenge of identifying massive amounts of data,artificial intelligence plays an increasingly important role in the field of pulsar candidate identification.Therefore,improving the identification accuracy of pulsar candidates based on deep learning has also become a significant challenge.This paper designs a joint model to identify candidates collected from the data of the fivehundred-meter aperture spherical radio telescope(FAST).This joint model utilizes ConvNeXt and DenseNet networks to perform feature learning on the two-dimensional and one-dimensional data of the candidates,respectively,and then uses a fully connected layer to discriminate the final classification results.In testing with the FAST dataset,the model achieves a recall rate of 94.8%and a precision rate of 95.7%,showing improvements in both recall and precision compared to the V-CCNN and H-CCNN met hods.

关 键 词：脉冲星筛选 ConvNeXt网络 DenseNet网络联合模型 

分 类 号：G45[文化科学—教育学]