多尺度轻量化CNN在SAR图像地物分类中的应用

孙盛, 蒙芝敏, 胡忠文, 余旭

Application of multi-scale and lightweight CNN in SAR image-based surface feature classification

SUN Sheng, MENG Zhimin, HU Zhongwen, YU Xu

表4 不同模型在GDUT-Nansha数据集上的对比

Tab.4 Comparison of different models on GDUT-Nansha data set(%)

模型 PA mPA IoU mIoU 植被 建筑物 水体 道路 FCN 87.5 78.3 72.8 80.1 89.7 30.3 68.2 SegNet 87.1 79.8 72.9 79.5 88.8 32.9 68.5 U-Net 88.0 80.0 73.3 80.5 90.5 34.1 69.6 DeepLabV3+ 87.1 80.2 70.5 79.7 89.1 37.0 69.1 ENet 85.7 77.7 69.3 76.7 88.8 29.4 66.1 ENet-CSPP 88.6 81.5 74.4 81.8 90.3 37.3 71.0