Simulating Infrared Facial Temperature for Recognition with a Heat Transfer-Inspired Model

Ambient temperature variations significantly affect facial temperature distribution, posing challenges for thermal infrared face recognition systems. To address this issue, this study proposes an infrared image augmentation method based on simulated facial temperatures. The method employs a heat transfer model to characterize the impact of ambient temperature changes on facial thermal patterns. To enhance the model’s fitting capability, the expression of thermal conductivity is processed nonlinearly. By fitting this parameter, the model can estimate facial temperature distributions under various external conditions, thereby augmenting the training dataset. Comparative experiments demonstrate that the proposed method improves the accuracy of temperature estimation and enhances recognition performance, especially in scenarios with limited training data.