ABSTRACT
Aurora phenomenon is caused by charged particles from solar-wind colliding with atmospheric gas molecules, which is the most visible manifestation of the Sun's influence on the Earth in high-latitude area. Detection and retrieval of auroral events possessing certain space structures and temporal variations is the most important means for aurora study, but is basically done by human vision so far. Because of the great variety in morphological and motion characters, auroral event is difficult to define and represent. In this paper, we propose an improved spatio-temporal transformer network to represent auroral event based on all-sky auroral images observed years 2003-2009 at Yellow River Station (YRS). Specifically, a context encoder is introduced to spatio-temporal transformer network architecture to leverage the spatial and temporal information, with the consideration of uncertain rate of aurora change. The detected diffusion auroral events are consistent with human visual perception, where the accuracy is improved by 1.2% than previous methods and the recall rate is 92.2%. This result can be applied in the detection of large-scale auroral events and improves the efficiency of aurora study.
- Han D S , Chen X C , Liu J J , An extensive survey of dayside diffuse aurora based on optical observations at Yellow River Station[J]. Journal of Geophysical Research, 2015, 120(9):7447-7465.Google ScholarCross Ref
- Han Desheng , Hu Zejun , Chen Xiangcai , New progress in the study of solar-side aurora based on observations from the Arctic Yellow River Station[J]. Polar Research, 2018, 030(003):235-250.Google Scholar
- Syrjäsuo M T, Donovan E F. Diurnal auroral occurrence statistics obtained via machine vision[J]. Annales geophysicae. Copernicus GmbH, 2004, 22(4): 1103-1113.Google Scholar
- Hu Z J , Yang H , D Huang , Synoptic distribution of dayside aurora: Multiple-wavelength all-sky observation at Yellow River Station in Ny-lesund, Svalbard[J]. Journal of Atmospheric and Solar-Terrestrial Physics, 2009, 71(8-9):794-804.Google ScholarCross Ref
- Yang X , Gao X , Tao D , Shape-Constrained Sparse and Low-Rank Decomposition for Auroral Substorm Detection[J]. IEEE Trans Neural Netw Learn Syst, 2016, 27(1):32-46.Google ScholarCross Ref
- Zhang D, Wu W, Cheng H, Image-to-video person re-identification with temporally memorized similarity learning[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2017, 28(10): 2622-2632.Google ScholarDigital Library
- Fang H , Hu H Q , Yang H G , A comparative study of the cosmic noise absorption Keograms generated by two different quiet day curve techniques[J]. Chinese Journal of Geophysics, 2015, 58(1):1-11.Google Scholar
- Qian W , Liang J , Hu Z J , Spatial texture based automatic classification of dayside aurora in all-sky images[J]. Journal of Atmospheric and Solar-Terrestrial Physics, 2010, 72(5-6):498-508.Google ScholarCross Ref
- Zhong Y , Rui H , Ji Z , Aurora Image Classification Based on Multi-Feature Latent Dirichlet Allocation[J]. Remote Sensing, 2018, 10(2):233.Google ScholarCross Ref
- Yang X, Wang N, Song B, BoSR: A CNN-based aurora image retrieval method[J]. Neural Networks, 2019, 116: 188-197.Google ScholarDigital Library
- Kavukcuoglu K, Sermanet P, Boureau Y L, Learning convolutional feature hierarchies for visual recognition[J]. Advances in neural information processing systems, 2010, 23.Google Scholar
- Yan B, Peng H, Fu J, Learning spatio-temporal transformer for visual tracking[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. 2021: 10448-10457.Google Scholar
- Tirupattur P, Duarte K, Rawat Y S, Modeling multi-label action dependencies for temporal action localization[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2021: 1460-1470.Google Scholar
- Carion N, Massa F, Synnaeve G, End-to-end object detection with transformers[C]//European conference on computer vision. Springer, Cham, 2020: 213-229.Google Scholar
- He K, Zhang X, Ren S, Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778.Google Scholar
- Teed Z, Deng J. Raft: Recurrent all-pairs field transforms for optical flow[C]//European conference on computer vision. Springer, Cham, 2020: 402-419.Google Scholar
- Yang X, Gao X, Song B, Aurora image search with contextual CNN feature[J]. Neurocomputing, 2018, 281: 67-77.Google ScholarDigital Library
- Liu S, Li B, Fan Y Y, Facial attractiveness computation by label distribution learning with deep CNN and geometric features[C]//2017 IEEE International Conference on Multimedia and Expo (ICME). IEEE, 2017: 1344-1349.Google Scholar
- Szegedy C, Vanhoucke V, Ioffe S, Rethinking the inception architecture for computer vision[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 2818-2826.Google Scholar
- Vaswani A, Shazeer N, Parmar N, Attention is all you need[J]. Advances in neural information processing systems, 2017, 30.Google Scholar
- Zhou B, Khosla A, Lapedriza A, Learning deep features for discriminative localization[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 2921-2929.Google Scholar
Index Terms
- Detection of Diffuse Auroral Events Based on Spatio-Temporal Transformer
Recommendations
Unsupervised Feature Learning for Temporal Segmentation of Auroral Image Sequences
AIPR '21: Proceedings of the 2021 4th International Conference on Artificial Intelligence and Pattern RecognitionOptical observation of aurora can obtain extensive information about magnetosphere and solar-terrestrial electromagnetic activities. Temporal segmentation of auroral image sequences is the basis for detection and retrieval of auroral events and further ...
Detection of Diffuse Auroral Events Based on LSTM
AIPR '21: Proceedings of the 2021 4th International Conference on Artificial Intelligence and Pattern RecognitionDiscrete and diffuse aurora are important categories of auroras optically observed on the ground. Some of the mechanisms and regularities of diffuse auroras are still unclear. In this work, we formulate the automatic diffuse aurora detection to learn a ...
Auroral Oval Boundary Modeling Based on Deep Learning Method
IScIDE 2015: Revised Selected Papers, Part II, of the 5th International Conference on Intelligence Science and Big Data Engineering. Big Data and Machine Learning Techniques - Volume 9243Research on the location of the auroral oval is important to understand the coupling processes of the Sun-Earth system. The equatorward boundary and poleward boundary of the auroral oval are significant parameters of the auroral oval location. Thus ...
Comments