Image Retrieval by Visual Examples for Supporting Euratom Nuclear Safeguards Activities
This paper presents a proposal for the architecture of an image retrieval system supporting Safeguards inspectors in the analysis of surveillance images from nuclear plants. A time-lapse system records about 25000 images every 3 months. Scene change detectors identify as relevant between 500 and 1500 images. From those, only 100 to 150 images are of real interest. In this paper we are investigating the possibility of using image retrieval by visual example for isolating Safeguards relevant events and we propose a system architecture integrating video-surveillance systems and image retrieval techniques.
Bibliographic Reference: An oral report given at: 2nd European Workshop on Advanced Video-Based Surveillance Systems. Organised by: University of Kingston. Held in: Kingston (UK), 04 September (2001)
Record Number: 200013783 / Last updated on: 2001-09-18
Original language: en
Available languages: en