Capturing task knowledge for geo-spatial imagery

Geo-spatial image databases are employed in a wide range of applications, such as intelligence operations, recreational and professional mapping, urban and industrial planning, and tourism systems. Effective retrieval of relevant images from such digital libraries can employ knowledge about what an image contains, why image contents are important in a particular domain, and how specific images have been used for particular domain tasks. Approaches to annotation for multimedia information retrieval have typically focused on the first two types of knowledge; however, managing the knowledge implicit in using geo-spatial imagery to address particular tasks can be crucial for capturing and making the most effective use of organisational knowledge assets. We are developing case-based knowledgemanagement support for large geo-spatial image repositories that scaffolds task-based knowledge capture about a content-based sketch query mechanism. This paper describes our task-centric approach to image annotation and retrieval, and it presents our initial implementation of the approach.