Flue Cured Tobacco Area Extraction of Remote Sensing Image by Integrating Lightweight ASPP and U-Net

Table 1. The experimental setup
View table
View in Article
Table 1. The experimental setup
名称环境配置
CPU AMD A10-7300，内存为8 GB
图形处理器GPU Telas V100，内存为16 GB
编程语言 python3.5
深度学习框架 TensorFlow1.14.0+keras2.3.1
CUDA版本 9.0
CUDNN版本 7.6.5

Table 2. The parameters setting
View table
View in Article
Table 2. The parameters setting
算法优化器迭代次数/次学习率批次大小
FCN Adam 100 0.000 1 16
U-Net Adam 100 0.001 32
SegNet Adam 100 0.001 32
DeepLabv3+ Adam 100 0.001 32
本文方法 Adam 100 0.0001 32

Table 3. The comparison of different tobacco area segmentation methods
View table
View in Article
Table 3. The comparison of different tobacco area segmentation methods
算法 PA/% MIoU/% 参数量/MByte 训练时间/h 单图预测时间/ms
U-Net 92.9 83.1 95.0 0.57 25
FCN 91.2 76.9 19.1 0.22 12
DeepLabv3+ 92.4 82.6 209.4 1.02 40
SegNet 92.0 81.0 44.1 0.30 15
本文方法 93.7 84.1 95.2 0.58 26

Table 4. The results of ablation study
View table
View in Article
Table 4. The results of ablation study
算法 PA/% MIoU/% 参数量/MByte 训练时间/h 单图预测时间/ms
U-Net 92.9 83.1 95.0 0.57 25
方案一 93.1 83.2 95.0 0.57 25
方案二 93.6 84.1 95.5 0.59 27
本文方法 93.7 84.1 95.2 0.58 26

Tools

Get Citation

Copy Citation Text

Shufeng HAO, Yu GAO, Ping LIU, Yuang LI, Huadong ZHANG, Hongjie REN, Shuaijie TIAN, Wentao KOU. Flue Cured Tobacco Area Extraction of Remote Sensing Image by Integrating Lightweight ASPP and U-Net[J]. Spacecraft Recovery & Remote Sensing, 2024, 45(4): 139

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites