Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Assessment procedure with specific ROC curves for comparison of fusion engines

: Sartor, Timo; Scherer-Negenborn, Norbert; Michaelsen, Eckart; Jäger, Klaus

Preprint urn:nbn:de:0011-n-1835553 (219 KByte PDF)
MD5 Fingerprint: 169181fd8782fc2cac0173508a3dcd99
© 2011 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.
Erstellt am: 7.12.2011

Dunham, Darin ; Institute of Electrical and Electronics Engineers -IEEE-; International Society of Information Fusion -ISIF-:
14th International Conference on Information Fusion 2011. Proceedings : Chicago, Illinois, USA, 5 - 8 July 2011
Piscataway, NJ: IEEE, 2011
ISBN: 978-1-4577-0267-9
ISBN: 978-0-9824438-2-8
7 S.
International Conference on Information Fusion (FUSION) <14, 2011, Chicago/Ill.>
Konferenzbeitrag, Elektronische Publikation
Fraunhofer IOSB ()
fusion engine; track fusion; performance evaluation; assessment procedure; ROC curve

Fusion Engines in general have to handle manifold types of information from different sensors. In particular, in urban terrain such diverse sensor systems as e.g. electro optical cameras, thermal cameras, small ground radars, acoustical sensors, and chemical, biological, radiological and nuclear (CBRN) sensors can contribute information. Several approaches to the fusion of such possibly contradictory or affirming information are known. This contribution evaluates such fusion results by estimating the gain of information comparing it to the individual sensor results. Specific ROC curves are used as evaluation criterion. This procedure opens the way for a comparison of Fusion Engines in general. To this end the evaluation procedure has to be capable of handling the diverse interfaces for the ground truth and sensor data, as well as for the fusion results. Common to all of these interfaces is that all information has to be labeled by geographical location and time, but they can have additional variables such as e.g. features or classification results. Input data can be synthetic or real data. Different quantitative measures are derived for the ranking of such systems.