综合像素级和特征级的建筑物变化检测方法  

Change Detection Method for Buildings Based on Pixel-level and Feature-level

作　　者：张永梅[1,2] 李立鹏[1] 姜明[2] 刘海伟[1] 

机构地区：[1]北方工业大学信息工程学院,北京100144 [2]北京大学数学科学学院,北京100871

出　　处：《计算机科学》2013年第1期286-293,共8页Computer Science

基　　金：国家科技支撑计划项目(2012BAH04F00);北京市高校学术创新团队项目(PHR201007121);北京市图像信息处理与智能识别科研平台建设项目(PXM2012_014212_000024);北方工业大学重点研究计划项目资助

摘　　要：针对单独使用像素级变化检测或特征级变化检测对于高层建筑物检测精度低的问题,提出了一种结合像素级和特征级的建筑物变化检测方法。首先对多个时相的遥感图像进行基于比值法的像素级变化检测,得到包含建筑物变化的候选区域,在候选区域上再进行基于建筑物特征的变化检测。该方法首先利用基于Delaunay三角网约束的快速配准算法配准两个不同时相的多光谱图像,利用建筑物的变化会导致建筑物所在局部区域的纹理分布和色调发生变化的特点,提取对辐射差异和配准误差鲁棒的纹理和色调特征进行变化检测。实验结果表明,该方法可以有效提高建筑物变化检测正确率,降低虚检率。Aiming at the problem that only using pixel-level or feature-level change detection for high-rise building has low accuracy,a method combined pixel-level and feature-level change detection was presented.Detect changes of multi-temporal remote sensing image based on ratio method,obtain the candidate change regions of high-rise buildings,then detect changes in the candidate regions based on building feature.Firstly,a novel fast registration algorithm of constraint-based Delaunay triangulation is used to make registration of multi-spectral images with two different phase.The building will lead to changes of distribution and characteristics of color in the texture of the local area changes,so,robust texture and color characteristics for radiation and registration are extracted to make change detection.Experiment results show that the combination of pixel-level and feature-level of the building change detection method can effectively improve the accuracy,and reduce the false alarm.

关 键 词：建筑物 变化检测 评价指标 多光谱图像 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]