In this paper, thermogram was used to do human face recognition. The thermogram of human face was captured from infrared camera. Image processing technologies were used to preprocess the captured thermogram. Then several features were extracted from the processed thermogram. These features included the distribution of temperature statistics, the perimeter of the face, the length/width ratio of the smallest rectangle which contained the face, and the triangle formed from the three lowest temperature points of the two cheeks and the chin. A three-layer back-propagation neural network was used as the recognition tool. The samples were collected from 20 individuals, 16 of them with glasses. The total samples have 200 for without glasses and 96 for with glass. Different combinations of the extracted features were tested. It is shown that the recognition rates for without glasses and with glasses are 95% and 93%, respectively. The experimental results encourage the application of thermogram for human face recognition.
|Number of pages||6|
|Journal||Journal of Medical and Biological Engineering|
|Publication status||Published - 2002 Jun 1|
All Science Journal Classification (ASJC) codes
- Biomedical Engineering