顔面熱画像の空間周波数特徴による個人識別

抄録

Recently, we are concerned with security because picking and imitation on computer network are increasing. Then, the approach to individual identification from biometric is expected, because it can realize high reliability and it's a natural method.
This paper's proposition is to construct individual identification system with the facial skin thermogram (FST) which is robust about illumination condition. FST has no shadow with illumination because it catches infrared rays light and makes an image. Then, we extract individual feature from FST by image processing and examine evaluation method to identify individual. However, facial skin temperature is ruled by autonomous nerve system. Especially, nasal skin temperature is affected by a mental influence like stress and arousal level. And we limited identification object to the measurement of face image in the rest state.
We applied two dimensional discrete fourier transform (2DDFT) as a method of extracting individual feature, because we tried to convert the individual feature of FST into spatial frequency feature. Afterwards, we identified individual by pattern matching. Identification experiment was done in 30 registered patterns and 200 input patterns, consequently correct acceptance rate 91.0[%] and correct rejection rate 91.0[%] was able to be obtained. Moreover, we confirmed individual feature appeared in low frequency element of FST.