• Laser & Optoelectronics Progress
  • Vol. 56, Issue 16, 160101 (2019)
Peng Lu**, Peiqi Zou, and Guoliang Zou*
Author Affiliations
  • College of Information Technology, Shanghai Ocean University, Shanghai 201306, China
  • show less
    DOI: 10.3788/LOP56.160101 Cite this Article Set citation alerts
    Peng Lu, Peiqi Zou, Guoliang Zou. Typhoon Classification Model Based on Multi-Scale Convolution Feature Fusion[J]. Laser & Optoelectronics Progress, 2019, 56(16): 160101 Copy Citation Text show less
    Convolutional feature visualization
    Fig. 1. Convolutional feature visualization
    Flow chart of MS-TyCNN model
    Fig. 2. Flow chart of MS-TyCNN model
    Preprocessing of satellite cloud images
    Fig. 3. Preprocessing of satellite cloud images
    Network structural diagram of MS-TyCNN model
    Fig. 4. Network structural diagram of MS-TyCNN model
    Structure of spatial pyramid pooling layer (T21=1/21)
    Fig. 5. Structure of spatial pyramid pooling layer (T21=1/21)
    Partial samples of typhoon labels
    Fig. 6. Partial samples of typhoon labels
    Comparison of classification results of typhoon cloud map datasets. (a) Accuracy of test set; (b) loss of validation set
    Fig. 7. Comparison of classification results of typhoon cloud map datasets. (a) Accuracy of test set; (b) loss of validation set
    LabelLevelof typhoonMaximum windspeed /(m·s-1)
    Class 1Tropical storm≤24.4
    Class 2Severe tropical storm24.5-32.6
    Class 3Typhoon32.7-41.4
    Class 4Violent typhoon41.5-50.9
    Class 5Super typhoon≥51.0
    Table 1. Typhoon classification standard
    ModelTrain accuracyTest accuracy
    LeNet-50.86570.8559
    AlexNet0.95080.9432
    Hybrid Model[17]0.97140.9366
    SIFT +CNN[18]0.92720.9205
    MS-TyCNN1.00000.9988
    Table 2. Accuracies of different models on typhoon datasets
    DatasetNumber oftraining sampleNumber oftesting sampleClass
    MNIST420001000010
    CIFAR-10500001000010
    Table 3. Distribution of datasets
    DatasetLeNet-5AlexNetHybrid modelSIFT+CNNMS-TyCNN
    MNIST0.96370.98060.97250.98370.9814
    CIFAR-100.73200.84310.89260.87590.8763
    Table 4. Generalization of MS-TyCNN model
    Peng Lu, Peiqi Zou, Guoliang Zou. Typhoon Classification Model Based on Multi-Scale Convolution Feature Fusion[J]. Laser & Optoelectronics Progress, 2019, 56(16): 160101
    Download Citation