In this paper, the researchers propose a new approach to identify the physiological thermal infrared image characteristics. First, using color segmentation algorithm to extract the face from the background. Then out of the nose by the dynamic temperature changes, through the nostrils to locate and formalize the length and width of the face image. The hair, glasses, nose and beard neither conducive to the stability of recognition. In this study, the range will be below the nostrils as the regional scope of face recognition. And divide the thermal image into five temperature layers. Finally, the use of statistical methods is to identify the temperature layer of the feature points. The results can also be found in the capture of different faces, and the resulting physical characteristics are not the same. The researchers successfully used this feature for thermal imaging of human face recognition. The proposed identification method is feasible by the field-proven, and the identification of cross-match success rate is 95.6%.