Seminar za računarstvo i primenjenu matematiku, 23. april 2019.

• 22. April,  2019
• Komentari (0)

Naredni sastanak Seminara biće održan u utorak, 23. aprila 2019. u sali 301f Matematičkog instituta SANU sa početkom u 14:15.

Predavač: Ranko Petrović, Miloš Pavlović, Vlatacom Institute of High Technologies

Naslov predavanja: FUSION OF VISIBLE AND THERMAL IMAGERY FOR ROBUST FACE RECOGNITION SYSTEMS

Apstrakt: Visible light face recognition systems have been extensively studied: in controlled environments the recognition accuracy can reach more than 99%. Regardless of that, variations in lighting conditions, pose, facial expressions and occlusions result in performance degradation. Utilization of thermal band imaging sensors, with the aim to overcome the disadvantages of visible light face recognition systems, is becoming more and more popular in a wide range of security and safety applications. Nearly invariant to changes in illumination, thermal IR imagery provides the ability for recognition under all lighting conditions, including complete darkness. However, thermal images have several deficiencies. IR imaging is sensitive to changes in the environment`s temperature, variations in the heat facial patterns, presence of noise originating both from sensor and environment, and as the most important, IR is opaque to glass. In order to fuse information and take advantages from both spectra, this research provides an algorithm for fusion of thermal and visible imagery, improving, in such a way, face recognition performance and robustness. Also, in order to develop a robust face recognition system and to be operational with the existing visible light facial databases, this research provides a comparative analysis of the available approaches to thermal-to-visible face recognition, and the most important signal processing technologies behind them.

Keywords: Face recognition, visible light imagery, thermal imagery, sensor fusion

---