地表覆盖变化信息是许多应用所需要的重要和基础信息。如何有效地从多时相遥感图像中提取变化信息一直是遥感信息处理与应用中的关键问题之一。针对现有的遥感变化检测研究在地物类别时空转换信息提取和多信息融合方面存在的问题，本项目提出建立地物类别的多点时空转换概率计算新方法，以更有效地表达地物类别的复杂空间-时相转换关系，并建立考虑多源信息相关性的决策融合方法，在此基础上建立融合光谱、空间和时相信息的变化检测新方法，并针对城市扩展应用通过与现有方法的对比来评价和验证项目所提出的新方法。项目的研究成果可为遥感变化检测、多时相图像分析、决策级融合研究以及相关应用提供新的方法和思路。

Land cover change information is an important and fundamental information source required in many applications. How to effectively extract land cover change information has been one of key problems in remote sensing information processing and applications. Considering the issues in extraction of spatio-temporal class transition and multiple information fusion for remote sensing based change detection, this project proposes to develop new methods to calculate multi-point class transition probabilities in spatial and temporal domains, to better quantify complex spatio-temporal class transitions from multitemporal images. The project will then develop decision fusion methods taking into account correlation between multiple sources of information. Furthermore, novel change detection methods which effectively fuses spectral, spatial and temporal information will be developed. The quantitative evaluation and validation will be conducted by comparing the proposed methods and existing methods in urban expansion analysis. The methods and results presented in the project will provide new ways for change detection, multitemporal analysis and decision fusion, as well as many relevant applications.