PNNL

Abstract

Data fusion (or information integration) from different sources is required for many technical fields, including epidemiology, medicine, biology, business, and military applications such as intelligence. Data fusion may involve integration of spectral content with imaging, text, and many other observations or with sensor data. Therefore, any software system for information integration should be extendable from one application to another. This paper describes some specific instances of data fusion, and shows how the fusion process can be viewed as multi-path and multi-stage mappings. These mappings can be organized from a software perspective into parallel development and management of four tools libraries: structure tools for data-fusion structure specifications, analytic tools for mapping methodologies, visualization tools for human and machine interface, and evaluation tools for evaluating fusion outcomes. Such software architecture will ensure the extendibility of the information integration system.