Thursday, May 2025
04:10 PM - 04:30 PM
Room: LL20A
Session: Artificial Intelligence for Automotive Displays and HMI Technologies
Improvement of Image Quality of Infrared Camera Behind LCD Screen and Its Application in DMS
Description:
Driver monitor system (DMS) is an essential configuration for safe driving. In order to miniaturize the car frame and maximize the interior space, we integrated the infrared camera behind the liquid crystal display (LCD). The infrared light source outside the screen shines on the driver’s face, and the reflected infrared light is received by the infrared camera behind the LCD screen. Compared to the image received by an unobstructed camera, the image it receives is very blurry. To solve this problem, we consider the LCD screen as composed of thin films with different media, and the image becomes increasingly blurry with each layer of thin film. A layer of thin film is similar to a layer of fog. The image after each layer of thin film is processed by defog and guided filtering, and captured by the infrared camera behind the LCD screen becomes clear layer by layer. Afterwards, we use a cascaded regressor for facial feature alignment. After multi-layer defogging, the facial recognition ratio has increased by nearly 70% compared to before. In addition, we propose two parameters representing abnormal head states based on two-dimensional facial images: HTR and HTG, which can sensitively reflect head turning and tilting actions.