CTIT University of Twente
Research Business & Innovation About CTIT Research Calls Looking for a job? Intranet

Face recognition on basis of non-calibrated caputered images

Face recognition on basis of non-calibrated caputered images

Project Manager: Dr. ir. R.N.J. Veldhuis
Faculty of Electrical Engineering, Mathematics and Computer Science
Tel.: +31 53 489 2838
Email: r.n.j.veldhuis@utwente.nl


The number of surveillance cameras at public places and shops is increasing, thus also the cases in which Closed Circuit Television images or captured images are used as evidence. In these cases possible ‘identification’ (who is the offender) or ‘authentication’ (is the suspect the offender) is taking place on basis of morphologic characteristics of the person, such as height, or more detailed characteristics, such as face or even ears. One of the problems is that we are faced with ‘non-calabrated’ or ‘uncontrolled’ images, images of different cameras and camera systems, of which the position and settings are different. This results in images of different quality and from different camera perspectives. The current techniques (automatic and manual) use only one or a few ‘best fitting’ images from a serie of images, thus quite a lot of less suitable material is neglected with the analysis. Use of all material could improve de evidence of surveillance images. To optimally use the information of images from different angles, or from moving persons, new 3D techniques are needed. Image sequences from videos can then be used for the reconstruction of a 3D model of persons on the images. The photographic methods for 3D reconstruction and the analisation of images not only have to provide an evidential verification, but should also be adaptable to material of different quality.

Project duration: 4 years / 2009-2012
Project budget: 905 k-€
Number of person/years: 2 fte/yr
Project Coordinator: NFI
Participants: NFI, UT
Project budget CTIT: 750 k-€
Number of person/years CTIT: 1.5 fte/yr
Involved groups: Signals & Systems (SAS)