基于改进的空洞卷积UNet网络提取林地信息方法

Research on forest information extraction based on improved dilated convolutional UNet network

为提升大范围林地信息提取的智能化程度及效率,提出了一种基于残差连接的空洞卷积UNet网络(RCD-UNet网络),将残差连接双卷积模块与传统UNet网络进行有机结合以提升模型性能,并以GF-2号高分辨率卫星遥感影像为数据源,开展了南京林地信息提取方法研究。结果表明:引入空洞金字塔池化层(ASPP)模块能够增强模型对上下文的感知能力,林地信息提取的总体精度为95.44%,Kappa系数为82.48%,满足高效、准确提取森林资源空间结构信息的需求,为森林资源管理与调查提供了技术支撑。

To improve the intelligence and efficiency of large-scale forest information extraction,a residual connected dilated convolutional UNet(RCD-UNet)network has been proposed.The residual connected dual convolutional module is organically combined with the traditional UNet network to improve model performance.An experiment on forest extraction in Nanjing has been conducted using high-resolution satellite remote sensing images of Gaofen-2 as the data source.The results indicate that the proposed method can enhance the model′s ability to perceive the context by introducing an atrous spatial pyramid pooling module.The overall accuracy of forest extraction is 95.44%,and the Kappa coefficient is 82.48%,which meets the requirements for efficient and accurate extraction of forest resource spatial structure information,providing technical support for forest resource management and investigation.