基于高分辨率影像的农业大棚信息提取  

Information Extraction of Agricultural Greenhouse Based on High Resolution Image

作　　者：刘森 马丙成 LIU Sen;MA Bingcheng(Bei Fang Investigation,Design&Research Co.,Ltd.,Tianjin 300222,China)

机构地区：[1]中水北方勘测设计研究有限责任公司,天津300222

出　　处：《现代信息科技》2023年第22期146-149,共4页Modern Information Technology

摘　　要：农业大棚在农业发展中具有举足轻重的地位,大棚提取对于农业的可持续发展和环境治理尤为重要。现有方法很难获得大棚的精确边界。文章通过对比6种常见语义分割模型,探究空洞卷积及空间注意力机制对大棚提取结果的影响。以安徽省宿州市萧县作为研究区域,使用吉林1号卫星遥感影像创建一个新的农业大棚数据集。SA_UNet获得了94.16%的准确率,明显优于主流语义分割网络模型结果,并且可以提取边界更精确的大棚结果。Agricultural greenhouses play an important role in agricultural development.The extraction of greenhouses is crucial to the sustainable development of agriculture and environmental governance.Existing methods are difficult to obtain the precise boundary of the greenhouse.Therefore,by comparing six common semantic segmentation models,this paper explores the impact of dilated convolution and spatial attention mechanism on the results of greenhouse extraction.Taking Xiaoxian County,Suzhou City,Anhui Province as the research area,a new agricultural greenhouse dataset is created using Jilin-1 satellite remote sensing images.SA_UNet achieved an accuracy rate of 94.16%,which is significantly better than the results of mainstream semantic segmentation network models,and can extract greenhouse results with more precise boundaries.

关 键 词：高分辨率影像 农业大棚 深度学习 语义分割 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]