【摘要】 从高分辨率光学遥感影像中提取地物信息受到广泛关注,道路提取是其中一项重要任务。设计了一种基于残差网络结构的深度学习神经网络模型,并将该模型与FCN网络模型和U-Net网络模型输出结果进行了对比。结果表明,本文方法不仅在F₁得分和IOU上均优于其他两个模型,并且在提取的道路平滑度和图像噪声处理方面也具有一定的优势。更多还原

【Abstract】 Extracting mapinformation from high-resolution optical remote sensing images has received widespread attention,and road extraction is one of the important tasks.This paper designed a deep learning neural network model based on residual network,and compared it with FCN network model and U-Net network model.Experimental analysis showed that the method in this paper was superior to the other two models in comparison of extraction results,both in F₁ score and IOU（Intersection Over Union）,and has certain advantages in road extraction smoothness and noise processing.更多还原