耦合卷积神经网络与注意力机制的无人机摄影测量果树树冠分割方法

何海清, 周福阳, 陈敏, 陈婷, 官云兰, 曾怀恩, 魏燕

Fruit Tree Canopy Segmentation by Unmanned Aerial Vehicle Photogrammetry Coupled on Convolutional Neural Network and Attention Mechanism

HE Haiqing, ZHOU Fuyang, CHEN Min, CHEN Ting, GUAN Yunlan, ZENG Huaien, WEI Yan

表2 不同网络模型的2D和2.5D数据树冠分割定量评价结果

Tab. 2 Quantitative evaluation results of 2D and 2.5D data crown segmentation for different network models (%)

方法 数据集 OA F1 mIoU 区域生长 2D 80.11 77.12 74.94 分水岭算法 2D 87.38 82.60 80.45 FCN 2D 95.29 88.71 86.97 2.5D 95.19 90.75 88.39 BiseNet-V2 2D 95.45 89.13 87.40 2.5D 95.64 91.23 89.21 PSPNet 2D 96.85 92.47 91.05 2.5D 97.09 94.41 92.94 EfficientNet-V2 2D 96.89 92.80 91.47 2.5D 96.75 93.99 92.16 DANet 2D 96.90 92.57 91.17 2.5D 97.15 94.73 93.08 DeepLab-V3 2D 97.15 93.26 91.85 2.5D 97.06 94.56 92.88 CSwin-Tiny 2D 97.08 93.29 91.96 2.5D 97.26 94.35 92.95 CNNAMNet 2D 97.20 94.62 93.03 2.5D 97.57 95.49 94.05