Application of thermal imaging for human face recognition

• Autorzy: Kobel J. , Suchwałko A. , Podbielska H.
• Języki publikacji: EN

Abstrakty

EN

In this paper the possibility of exploiting thermal images of human faces and their computer aided analysis for person recognition is presented. The statistical pattern recognition methods realizing the classification and identification of human faces thermograms are applied. For thermogram recording the spectral range of 8–13 mm was chosen. The applied procedure is a two step scheme: dimensionality reduction of the data based on the PCA (principal component analysis) approach and cluster analysis. Five hierarchical methods of cluster analysis and five different distance metrics were used for classification. The recognition was carried out for six different sizes of face images: 85×85, 50×50, 25×25, 13×13, 7×7 and 3×3 pixels. It was shown that the cluster analysis by the Ward method gave the best results, namely the highest recognition accuracy (RA). The RA depends on the size of an analyzed image, as well. It decreases with size reduction.

• Wydawca: Oficyna Wydawnicza Politechniki Wrocławskiej
• Czasopismo: Optica Applicata
• Rocznik: 2002
• Tom: Vol. 32, nr 4
• Strony: 653--664
• Opis fizyczny: bibliogr. 14 poz.

Twórcy

autor

Kobel J.

autor

Suchwałko A.

autor

Podbielska H.

• Bio-Optics Group, Institute of Physics, Wrocław University of Technology, Wybrzeże Wyspiańskiego 27, 50-370 Wrocław