Typhoon Class Prediction Method Based on Physical Constraints and Cloud Map Generation

Zongsheng Zheng; Wenhuan Zhou; Zhenghan Wang; Meng Gao; Zhijun Huo; Yuewei Zhang

doi:10.3788/LOP240513

Laser & Optoelectronics Progress, Volume. 61, Issue 24, 2428005(2024)

Typhoon Class Prediction Method Based on Physical Constraints and Cloud Map Generation

Zongsheng Zheng¹, Wenhuan Zhou^1、*, Zhenghan Wang¹, Meng Gao¹, Zhijun Huo¹, and Yuewei Zhang²

Author Affiliations

¹Shanghai Ocean University, School of Information Technology, Shanghai 201306, China

²Guangzhou Meteorological Satellite Ground Station, Guangzhou 510640, Guangdong , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(11)

Fig. 1. Overall architecture of CPGANTyphoon model

Download full size

Fig. 2. Generator unit structure

Download full size

Fig. 3. PDE physical unit

Download full size

Fig. 4. Process image of typhoon 11, 2021 in the South Pacific. (a)-(h) Images of typhoon at 12 h intervals

Download full size

Fig. 5. Example of prediction results. (a1-4) Real images after preprocessing; (b1-4) ConvLSTM; (c1-4) GAN-LSTM; (d1-4) PhyDNet; (e1-4) CPGANTyphoon

Download full size

Fig. 6. Example of fuzzy c-mean segmentation results for typhoon images. (a) Real image; (b) segmented image

Download full size

Fig. 7. Example of fuzzy c-mean segmentation results. (a1‒2) ConvLSTM; (b1‒2) GAN-LSTM; (c1‒2) PhyDNet; (d1‒2) CPGANTyphoon

Download full size

Table 1. Criteria for classifying typhoon

View table

Table 1. Criteria for classifying typhoon

Typhoon classification	Label	Maximum wind speed （10 min average）
Typhoon classification	Label	kt	m /s	km /h
Tropical depression （TD）	Class 1	-33	-17	-62
Tropical storm （TS）	Class 2	34‒47	18‒24	63‒88
Strong tropical storm （STS）	Class 3	48‒63	25‒32	89‒118
Typhoon （TY）	Class 4	64‒84	33‒43	119‒156
Strong typhoon （STY）	Class 5	85‒104	44‒53	157‒192
Super typhoon （Super TY）	Class 5	105‒	54‒	193‒

Table 2. Model index comparison of model on typhoon image dataset
View table
Table 2. Model index comparison of model on typhoon image dataset
Model SSIM PSNR MSE×10 MAE×10
ConvLSTM 0.730 17.87 0.154 0.976
GAN-LSTM 0.628 25.57 0.366 1.509
PhyDNet 0.584 19.50 0.566 2.182
CPGANTyphoon 0.916 30.36 0.012 0.250

Table 3. Comparison of accuracy of different models in predicting image fuzzy c-means segmentation results
View table
Table 3. Comparison of accuracy of different models in predicting image fuzzy c-means segmentation results
Model FCM accuracy
ConvLSTM 0.887
GAN-LSTM 0.893
PhyDNet 0.878
CPGANTyphoon 0.981

Table 4. Comparison of typhoon class prediction accuracy of different model prediction images
View table
Table 4. Comparison of typhoon class prediction accuracy of different model prediction images
Model Accuracy
ConvLSTM 0.780
GAN-LSTM 0.751
PhyDNet 0.833
CPGANTyphoon 0.985

Tools

Get Citation

Copy Citation Text

Zongsheng Zheng, Wenhuan Zhou, Zhenghan Wang, Meng Gao, Zhijun Huo, Yuewei Zhang. Typhoon Class Prediction Method Based on Physical Constraints and Cloud Map Generation[J]. Laser & Optoelectronics Progress, 2024, 61(24): 2428005

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites