Classification and Estimation of Typhoon Intensity from Geostationary Meteorological Satellite Images Based on Deep Learning

In this paper, a novel typhoon intensity classification and estimation network (TICAENet) is constructed to recognize intensity. The TICAENet model based on the LeNet-5 model, which uses weight sharing reduce number of training parameters, VGG16 replaces large convolution kernel with multiple small kernels improve feature extraction. Satellite cloud images typhoons over Northwest Pacific Ocean South China Sea from 1995–2020 are taken as samples. results show that accuracy 10.57% higher than model; 97.12%, precision 97.00% for tropical storms, severe storms super typhoons. mean absolute error (MAE) root square (RMSE) samples in 2019 4.78 m/s 6.11 m/s, 18.98% 20.65% statistical method, respectively. Additionally, takes less memory runs faster due kernels. proposed performs better other methods. general, can be used accurately classify estimate maximum wind speed by extracting features geostationary meteorological satellite images.