ZHANG Zhuoran,ZHANG Qian,SONG Zhi,et al.Research on Weather Image Recognition based on Residual Swin Transformer[J].Journal of Chengdu University of Information Technology,2023,38(06):637-642.[doi:10.16836/j.cnki.jcuit.2023.06.003]
基于残差Swin Transformer的天气图像识别技术研究
- Title:
- Research on Weather Image Recognition based on Residual Swin Transformer
- 文章编号:
- 2096-1618(2023)06-0637-06
- 关键词:
- 天气现象; 图像识别; 深度学习; Swin Transformer
- Keywords:
- weather phenomena; image recognition; deep learning; Swin Transformer
- 分类号:
- TP391.4
- 文献标志码:
- A
- 摘要:
- 人类活动经常受到天气条件的影响，基于图像的自动天气识别在实际应用中具有重要意义。然而现有方法均使用卷积神经网络，未能有效地利用图像的全局信息和像素点之间长距离的依赖关系，且识别的天气类型较少，识别精度较低。为解决这些问题，尝试将视觉Transformer应用到天气识别领域，同时提出一种基于残差Swin Transformer的模型，并使用先进的优化器Ranger来提高天气识别的正确率。该模型在包含11种天气现象的公开数据集WEAPD上进行验证，实验结果表明，其整体性能优于其他先进的识别网络，识别正确率达到93.6%，可为天气图像识别和天气预报研究提供参考。
- Abstract:
- Human activities are often affected by weather conditions, and automatic weather recognition-based image is of great importance in practical applications. However, existing methods all use convolutional neural networks, which fail to effectively utilize the global information of images and the long-distance dependency between pixel points, and recognize fewer weather types with low recognition accuracy. To solve these problems, we try to apply the visual Transformer to the field of weather recognition, and also propose a model based on the residual Swin Transformer and use the advanced optimizer Ranger to improve the weather recognition rate. The model is validated on WEAPD, a publicly available dataset containing 11 weather phenomena, and the results show that its overall performance is better than other advanced recognition networks, with a 93.6% correct recognition rate. It can benefit the research of weather image recognition and weather forecasting.
参考文献/References:
[1] Lu C,Lin D,Jia J,et al.Two-class weather classifica-tion[C].Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2015:3718-3725.
[2] Song H,Chen Y,Gao Y.Weather condition recognition based on feature extraction and K-NN.Berlin:Springer,2014:199-210.
[3] Krizhevsky A,Sutskever I,Hinton G E.Imagenet classification with deep convolutional neural networks[J].Communications of the ACM,2017,60(6):84-90.
[4] Lin D,Lu C,Huang H,et al.RSCM:Region selection and concurrency model for multi-class weather recognition[J].IEEE Transactions on Image Processing,2017,26(9):4154-4167.
[5] Zhao B,Li X,Lu X,et al.A CNN–RNN architecture for multi-label weather recognition[J].Neurocomputing,2018,322: 47-57.
[6] Wang C,Liu P,Jia K,et al.Identification of weather phenomena based on lightweight convolutional neural networks[J].CMC-COMPUTERS MATERIALS & CONTINUA,2020,64(3):2043-2055.
[7] Tan L,Xuan D,Xia J,et al.Weather Recognition Based on 3C-CNN[J].KSII Transactions on Internet and Information Systems(TIIS),2020,14(8):3567-3582.
[8] Xiao H,Zhang F,Shen Z,et al.Classification of Weather Phenomenon From Images by Using Deep Convolutional Neural Network[J].Earth and Space Science,2021,8(5):e2020EA001604.
[9] Dosovitskiy A,Beyer L,Kolesnikov A,et al.An image is worth 16x16 words:Transformers for image recognition at scale[C].International Conference on Learning Representations,2021.
[10] 刘文婷,卢新明.基于计算机视觉的Transformer研究进展[J].计算机工程与应用,2022,58(6):1-16.
[11] Liu Z,Lin Y,Cao Y,et al. Swin transformer: Hierarchical vision transformer using shifted windows[C].Proceedings of the IEEE/CVF International Conference on Computer Vision.2021:10012-10022.
[12] He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[C].Proceedings of the IEEE conference on computer vision and pattern recognition.2016:770-778.
[13] Huang G,Liu Z,Laurens V,et al.Densely Connected Convolutional Networks[C].IEEE Computer Society.IEEE Computer Society,2016.
[14] Wright L,Demeure N.Ranger21:a synergistic deep learning optimizer[J].arXiv preprint arXiv:2106.13731,2021.
[15] Liu L,Jiang H,He P,et al.On the variance of the adaptive learning rate and beyond[C].ICLR,2020.
[16] Zhang M,Lucas J,Ba J,et al.Lookahead optimizer:k steps forward,1 step back[J].Advances in neural information processing systems,2019,32.
[17] Wang C,Liu P,Jia K,et al.Identification of weather phenomena based on lightweight convolutional neural networks[J].CMC-COMPUTERS MATERIALS & CONTINUA,2020,64(3):2043-2055.
相似文献/References:
[1]陈 留,杨笔锋,谢 欢,等.基于纹理和SVM的地面凝结现象观测方法研究[J].成都信息工程大学学报,2022,37(06):622.[doi:10.16836/j.cnki.jcuit.2022.06.002]
CHEN Liu,YANG Bifeng,XIE Huan,et al.A Texture-based SVM Observation of Ground Condensation Phenomenon[J].Journal of Chengdu University of Information Technology,2022,37(06):622.[doi:10.16836/j.cnki.jcuit.2022.06.002]
备注/Memo
收稿日期:2022-11-20
基金项目:四川省科技厅资助项目(2021005); 四川省重点实验室科技发展基金资助项目(2018-青年-11)
通信作者:何嘉.E-mail:hejia@cuit.edu.cn