電子情報通信学会総合大会講演要旨
D-12-14
A Preliminary Study on Human Attribute Classification in Thermal Image
◎Brahmastro Kresnaraman・Daisuke Deguchi・Ichiro Ide・Hiroshi Murase(Nagoya Univ.)・Tomokazu Takahashi(Gifu Shotoku Gakuen Univ.)
For security surveillance with normal cameras, illumination is key. However, during nighttime or in poorly lit areas where illumination is far from ideal, cameras that work in thermal spectrum can be a better option. Some visual attributes of a person are easily distinguishable in this spectrum. This research focuses on classifying these visual attributes in thermal images. Experiments are conducted to compare classification performances by SVMs in both thermal and visible spectra. Results show the average F-Score in the thermal spectrum is higher, therefore classification of selected attributes is better in thermal spectrum.