【作者】 刘琳洁； 王博； 郭翠翠； 贺成龙； 刘文楷；

【Author】 LIU Linjie;WANG Bo;GUO Cuicui;HE Chenglong;LIU Wenkai;School of Artificial Intelligence and Computer Science, North China University of Technology;Space Star Technology Co., Ltd.;National Key Laboratory of Spatial Information System and Integrated Application;

【机构】 北方工业大学人工智能与计算机学院； 航天恒星科技有限公司； 空间信息体系与融合应用全国重点实验室；

【摘要】 遥感图像变化检测在城市规划和灾害评估等领域具有重要应用。然而,在视角变化、地物形状复杂的场景中,传统方法既容易受到伪变化干扰,也难以精确刻画变化区域的边界。针对上述问题,提出一种基于边界信息引导的双时相遥感图像变化检测方法。设计交换式双编码器-单解码器结构,有效减缓视角差异引起的特征干扰。利用Sobel算子获取精准的地物边界信息,增强网络对变化区域空间细节的捕捉能力。使用SYSU-CD数据集进行训练和验证。实验结果表明,所提方法在双时相遥感图像变化检测任务中表现出色,精确率达到87.63%,召回率达到83.75%,F1分数达到85.39%,IoU达到79.08%,验证了提出的方法在视角差异与复杂地物形状场景中的有效性。更多还原

【Abstract】 Remote sensing image change detection has important applications in fields such as urban planning and disaster assessment. However, in scenes with changing viewpoints and complex shapes of objects, traditional methods are easily disturbed by pseudo-changes and have difficulty accurately depicting the boundaries of the changed areas. To address the above problems, this paper proposes a change detection method for dual-temporal remote sensing images based on boundary information guidance. First, a switching dual encodersingle decoder structure is designed to effectively mitigate feature interference caused by viewpoint differences.Second, the Sobel operator is used to obtain accurate object boundary information and enhance the network’s ability to capture spatial details of the changed area. This paper uses the SYSU-CD dataset for training and verification. Experimental results show that the proposed method performs well in the task of change detection in dual-temporal remote sensing images, with a precision of 87.63%, a recall of 83.75%, an F1 score of 85.39%,and an IoU of 79.08%, which verifies the effectiveness of the proposed method in scenes with viewpoint differences and complex object shapes.更多还原