变化检测技术越来越多地应用于城市遥感分析和应用领域,但目前城市变化检测的研究主要基于中低空间分辨率的遥感数据,使用的方法也主要是像元直接比较法或者是分类后比较法。提出一种基于变化向量分析和相似度验证相结合的变化检测方法,应用高空间分辨率影像来快速实现城市建筑物、街道等目标的自动变化检测。并详细阐述了变化目标的提取以及验证的方法和过程,其结果真实地反映了地面目标的实际变化程度和类型。

As human activities expanding, land use and land cover change very quickly at different scales all over the world. Remote sensing becomes a major tool to acquire information of LUCC. In recent years, the continuing development of remote sensing technology provides us a large amount of remote sensing data at high spatial, spectral and temporal resolutions. Advances in remote sensing science and diversity in high resolution data hold great promise for improving the precision of information extraction and change detection, which also make change detection of land use and land cover at different scales from global scale to local scale more difficult. However, conventional remote sensing change detection techniques are inefficient due to the high spatial heterogeneity of inner objects in the image, more texture, more details and clear edges. Moreover, the requirement for real time and effective change detection methods and large size of high spatial resolution imagery cells for development of more automatic techniques of change detection. The method of change detection based on integration of change vector analysis and similarity calibration is presented for high spatial resolution data. It can be used to detect the change of building and street quickly and automatically. In this paper, we present details of the method of change object extraction and verification The methods are illustrated with an airborne linear scanner sensor image over the suburb of Tokyo city, Japan. The result of change detection will be compatible to complexity and fuzzy degree of change of object in high spatial resolution imagery at different times, which is distinguishable to the results using conventional change detection, in which the result only provide "change" and "no-change". The experimental results suggest that change detection based on object similarity calibration is more reliable, efficient than post classification change detection using high spatial resolution imagery.