中国科学院大学学报2024，Vol.41Issue(3)：375-386,12.DOI:10.7523/j.ucas.2023.060
基于改进U-Net模型的农业大棚遥感提取方法
Remote sensing extraction method of agricultural greenhouse based on an improved U-Net model
摘要
Abstract
The agricultural greenhouse is a kind of agricultural facility,which is divided into transparent and non-transparent according to the surface transmittance.The large-scale statistics of agricultural greenhouses are of great significance to the survey of agricultural facilities,the formulation of agricultural policies,and the planning of county economic development.Aiming at the problem that manual statistics are time-consuming and laborious,this paper utilizes the convolutional neural network to extract agricultural greenhouses information from high-resolution remote sensing images.To solve the problems of insufficient semantic information extraction in remote sensing images and insufficient utilization of edge information of the U-Net model,this paper proposes the following improvements:1)The semantic segmentation task is optimized,and ConvNeXt and attention mechanism is utilized to extract deep semantic information of agricultural greenhouses in remote sensing images.2)The edge detection task is introduced,and the gated convolution layer and concate operation are used to fuse the semantic features of the encoder and the image gradient output by the decoder,and then the edge information is combined to optimize the segmentation results.After testing,the improved model can extract both transparent and non-transparent agricultural greenhouses information at the same time and the recognition effect is good,which is greatly improved compared with the traditional method.关键词
U-Net/Google影像/多任务学习/农业大棚信息提取Key words
U-Net/Google images/multi-task learning/information extraction of agricultural greenhouses分类
农业科技引用本文复制引用
王寅达,彭玲,陈德跃,李玮超..基于改进U-Net模型的农业大棚遥感提取方法[J].中国科学院大学学报,2024,41(3):375-386,12.基金项目
全球能源互联网集团有限公司科技项目(SGGEIG00JYJS2100032)资助 （SGGEIG00JYJS2100032）
