摘要
精准农业是指信息技术与农业生产全面结合的新型农业。农作物信息和数据是精准农业中最核心的内容,通常使用无人机遥感技术获取农作物信息和数据。文中首先利用无人机采集农田数据,并根据地面参考数据以及相关资料,利用人工标注构建无人机遥感农田样本数据集,利用数据增强的策略扩充样本数据集;其次,提出一种改进的U-Net模型,即自注意力U-Net模型,将其应用于农作物分类。所提模型在传统U-Net的基础上加入自注意力机制,能够提高模型的特征学习能力以及泛化能力。使用所提方法在无人机遥感农田数据集上进行实验得出,与基线模型相比,所提模型能够提升农作物分类性能。
Precision agriculture is a new type of agriculture which combines information technology with agricultural production. Crop information and data are the core content of precision agriculture,and UAV remote sensing technology is usually used to obtain crop information and data. The UAV is used to collect farmland data,the manual annotation is used to construct the UAV remote sensing farmland sample dataset according to the ground reference data and related data,and the data augmentation strategy is used to expand the sample dataset. The improved U-Net model-self-attention U-Net model is proposed and applied to the crop classification. On the basis of traditional U-Net,the self-attention mechanism is added into the proposed model,which can improve the model′ s feature learning ability and generalization ability. Using the proposed method to experiment on UAV remote sensing farmland dataset,it is concluded that the proposed model can improve the performance of crop classification compared with the baseline model. The proposed method is used to conduct experiments on the UAV remote sensing farmland dataset,and the results show that in comparison with the baseline model,the proposed model can improve the crop classification performance.
作者
赵子宇
石刚
ZHAO Ziyu;SHI Gang(School of Information Science and Engineering,Xinjiang University,Urumqi 830049,China;Xinjiang Key Laboratory of Signal Detection and Processing,Xinjiang University,Urumqi 830049,China)
出处
《现代电子技术》
2023年第4期125-129,共5页
Modern Electronics Technique
基金
国家自然科学基金资助项目(62162059)
国家自然科学基金资助项目(12061072)。
关键词
精准农业
无人机
遥感技术
农田样本
数据增强
自注意力
U-Net
农作物分类
precision agriculture
UAV
remote sensing technology
farmland samples
data augmentation
self-attention
U-Net
crop classification
