基于深度学习的积雪覆盖区山地冰川识别研究

王晶晶,柯长青,陈军

Research on Identification of Snow-Covered Mountain Glacier based on Deep Learning

Jingjing WANG,Changqing KE,Jun CHEN

表4 基于不同backbone的U-Net网络冰川识别精度

Table 4 Classification accuracies of glaciers using U-Net network based on different backbones

模型名称 参数量/MB P R F1值 mIoU OA 运行时长s/轮 U-Net 29.60 0.896 2 0.867 6 0.881 7 0.853 7 0.937 8 264 MobileNetv2-Unet 2.82 0.837 8 0.906 2 0.870 7 0.844 4 0.935 7 232 VGG16-Unet 95.00 0.883 4 0.908 2 0.895 6 0.871 0 0.946 8 275 VGG19-Unet 115.00 0.894 7 0.904 6 0.899 6 0.875 2 0.948 4 291 Resnet18-Unet 84.90 0.857 0 0.920 0 0.887 4 0.862 7 0.943 8 324 Resnet34-Unet 123.00 0.870 2 0.899 8 0.899 8 0.858 9 0.941 4 392 Resnet50-Unet 69.80 0.843 2 0.914 0 0.877 2 0.851 6 0.939 0 407 EfficientNetB0-Unet 54.50 0.888 5 0.889 7 0.889 1 0.863 0 0.942 7 257 EfficientNetB7-Unet 296.00 0.882 8 0.871 0 0.876 9 0.848 9 0.936 0 604