Face Identification and Verification in Thermal Images

Surveillance has become a crucial role in recent years, owing to the rise in crime rates. Existing daytime surveillance research has improved performance by employing deep learning algorithms to recognize and track objects. However, due to low illumination and/or poor weather conditions, achieving the same performance for night vision are difficult. Face detection is a crucial task in surveillance. A face detection module based on merging of thermal and visual pictures is proposed. The encoder-decoder network of the fusion module extracts effective features from the given thermal and visual images using depth-wise convolution. By recording radiation information at a greater wavelength, the thermal face solves most of the limitations that visual faces have. The range of facial imageries that can be recognized is expanded by developing a method that recognizes thermal faces. Face recognition in thermal images is having high research scope, and it has been discovered that when thermal images are trained, it performs better.