河北工业科技2025，Vol.42Issue(3)：212-220,9.DOI:10.7535/hbgykj.2025yx03002
基于高分遥感影像的农作物面积分割研究
Research on crop area segmentation based on high-resolution remote sensing images
摘要
Abstract
To solve the problem of low accuracy of traditional image segmentation networks in crop area segmentation,a VGNet network model for feature fusion of parallel networks was proposed.Firstly,a VGNet network model with a parallel network structure was designed;Secondly,the VGNet network was utilized for the segmentation and extraction of crop areas;Finally,ablation experiments and comparative experiments were conducted based on the regional data set of wheat in Quyang County.The results show that the accuracy rate and average pixel precision(MPA)of the VGNet model in the task of crop area segmentation reach 94.57％and 94.13％respectively,which are increasing 6.88 and 8.72 percentage points compared to the traditional u-shaped network(UNet)model,increasing 1.72 and 2.23 percentage points compared to the high-resolution network(HRNet),and increasing 1.53 and 1.96 percentage points compared to the DeepLabV3+network,respectively.The VGNet network model can improve the accuracy of crop area segmentation and provide an effective method for remote sensing image segmentation.关键词
计算机图像处理/农作物面积分割/遥感图像/UNet模型/ViT模型Key words
computer image processing/crop area segmentation/remote sensing image/UNet model/ViT model分类
计算机与自动化引用本文复制引用
王建霞,刘文宇,王婵,吴长莉..基于高分遥感影像的农作物面积分割研究[J].河北工业科技,2025,42(3):212-220,9.基金项目
军民融合发展研究课题(HB24JMRH007) （HB24JMRH007）
