Seeing in the Dark: A Different Approach to Night Vision Face Detection with Thermal IR Images

Identifying faces in low-light situations can be a difficult feat because of the reduced visibility and substandard image quality. Conventional methods of face detection rely on visible light, which is insufficient in environments with low-light conditions. Our paper introduces a fresh approach to identifying faces in night vision. We make use of thermal infrared (IR) images to detect faces. Thermal IR images capture the thermal signatures of objects, which remain unaffected by low-light conditions, providing valuable information for accurate face detection. Our method utilizes a deep learning model trained on thermal IR images to detect faces in conditions with low lighting. To evaluate our approach, we tested it on a dataset of thermal IR images captured in different lighting scenarios and compared its performance with traditional face detection methods. The results of our experiments indicate that our proposed approach surpasses traditional face detection methods in low-light conditions, achieving high accuracy in detecting faces. Through a qualitative analysis of thermal IR images, we delved into the key factors that lead to our approach's success. Our findings show that the thermal signature of the face offers valuable insights that aid in precise face detection even in low-light environments. Furthermore, we assessed the effectiveness of our approach in detecting faces under different pose and expression variations, and the results indicate that our method is highly efficient in detecting faces in various pose and expression conditions. Our solution presents a novel and efficient method for detecting faces in low-light conditions using thermal infrared images, which has the potential to be utilized in various applications including surveillance, security and law enforcement. © 2023 Author.