Data models and algorithms for knowledge discovery in spatio-temporal data

最近的研究表明，原始空间数据量急剧增加。随着新的地球观测系统、全球定位系统支持的移动的设备以及无线地理传感器网络等其他新兴技术的引入，生成大量空间数据（特别是图像）的能力可能会进一步增强。这些趋势创造了一个全新的局面，我们生成空间数据（特别是图像）的能力已经超过了我们组织、处理和分析空间数据的能力。此外，前所未有的丰富的传感设备使我们能够随着时间的推移快速收集空间数据，从而产生大量的时空数据。因此，一个主要的新兴研究挑战是开发高效，可靠和强大的方法和范式，从这些数据中推断和发现高层次的知识，同时解决空间实体的复杂性和动态性。有鉴于此，解决的根本挑战是如何将遥感图像时间序列中的低级原始像素数据与未显式存储在图像中的高级定性时空知识联系起来。这种定性知识应该允许表示动态物理系统的所有方面（空间，时间和数量），并支持自动推理。本建议的目标是开发数据模型和算法的建模，抽象和定性知识发现的时空遥感图像数据。

Recent research points to a dramatic increase in the volume of raw spatial data. The ability to generate massive amounts of spatial data (in particular imagery) is likely to further intensify with the introduction of new earth observation systems, GPS-enabled mobile devices, and other emerging technologies such as wireless geo-sensor networks. These trends have created an entirely new situation, in which our ability to generate spatial data (and in particular imagery) surpasses our ability to organize, process, and analyze it. Furthermore, the unprecedented wealth of sensory devices has made it possible to rapidly collect spatial data over time, resulting in large amounts of spatio-temporal data. A primary emerging research challenge is therefore the development of efficient, reliable, and robust methods and paradigms for inferring and discovering high-level knowledge from such data, while addressing the complex and dynamic nature of spatial entities. In light of this, the fundamental challenge addressed is how to relate low-level raw pixel data in remotely sensed imagery time series to high level qualitative spatio-temporal knowledge which is not explicitly stored in the imagery. Such qualitative knowledge should allow representing all aspects of a dynamic physical system (space,time, and quantity), and support automated reasoning. The objective of this proposal is to develop data models and algorithms for modeling, abstraction, and qualitative knowledge discovery in spatio-temporal remotely sensed imagery data.