Towards a Unified Approach to Information Integration - A review paper on data/information fusion

Whitney, Paul D. ; Posse, Christian ; Lei, Xingye C.

Pacific Northwest National Laboratory (U.S.)

关键词: Visualization;    99 General And Miscellaneous//Mathematics, Computing, And Information Science;    Data Fusion;    Implementation;    Acoustics;   

DOI  :  10.2172/881949
RP-ID  :  PNNL-15428
RP-ID  :  AC05-76RL01830
RP-ID  :  881949

美国|英语

来源: UNT Digital Library

【 摘 要 】

Information or data fusion of data from different sources are ubiquitous in many applications, from epidemiology, medical, biological, political, and intelligence to military applications. Data fusion involves integration of spectral, imaging, text, and many other sensor data. For example, in epidemiology, information is often obtained based on many studies conducted by different researchers at different regions with different protocols. In the medical field, the diagnosis of a disease is often based on imaging (MRI, X-Ray, CT), clinical examination, and lab results. In the biological field, information is obtained based on studies conducted on many different species. In military field, information is obtained based on data from radar sensors, text messages, chemical biological sensor, acoustic sensor, optical warning and many other sources. Many methodologies are used in the data integration process, from classical, Bayesian, to evidence based expert systems. The implementation of the data integration ranges from pure software design to a mixture of software and hardware. In this review we summarize the methodologies and implementations of data fusion process, and illustrate in more detail the methodologies involved in three examples. We propose a unified multi-stage and multi-path mapping approach to the data fusion process, and point out future prospects and challenges.