结合掩码图像建模和对比学习的道路提取方法  

A Method for Road Extraction Using Masked Image Modeling and Contrastive Learning

作　　者：伍江江 李政宏 沙志超 陈浩 彭双 杜春 李军[1] Wu Jiangjiang;Li Zhenghong;Sha Zhichao;Chen Hao;Peng Shuang;Du Chun;Li Jun(College of Electronic Science and Technology,National University of Defense Technology,Changsha 410073,China)

机构地区：[1]国防科技大学电子科学学院,湖南长沙410073

出　　处：《系统仿真学报》2025年第4期922-932,共11页Journal of System Simulation

基　　金：国家自然科学基金(42471403,62106276)。

摘　　要：针对遥感影像道路提取的遮挡问题,提出了结合掩码图像建模和对比学习的道路提取方法,其模型训练过程包括掩码预训练阶段和对比训练阶段。掩码预训练阶段主要进行掩码图像重建,训练模型从部分区域被随机遮挡的图像恢复出整张图像;对比训练阶段主要对于预测错误以及置信度较低的区域进行对比学习,拉近相同类别对应的特征距离而增大不同类别特征间的距离。实验结果验证了方法的有效性和可用性。Aiming at the occlusion problem of road extraction from remote sensing images,a road extraction method combining MIM and CL is proposed,the model training process includes a masked pre-training stage and a contrast training stage.The masked pre-training stage mainly carries out mask image reconstruction,and trains the model to recover the whole image from some areas that are randomly occluded.The comparison training stage is mainly for the prediction error and low confidence regions to learn the comparison,to narrow the distance between the features of the same category and increase the distance between the features of different categories.The experimental results verify the effectiveness and usability of this method.

关 键 词：遥感影像 道路提取 掩码图像建模 对比学习 图像重建 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]