A new method for monitoring the state of drivers-occupant based on infrared 3D camera

In the field of transportation, such as aircraft and automobiles, certain states of the driver and occupants are important factors for safe driving. Monitoring the state, movement and behavior of pilots, car drivers and occupants can greatly reduce the occurrence of accidents. This study proposes a new method of driver-occupant state monitoring based on infrared 3D camera and deep learning to achieve dangerous state monitoring of drivers and occupants, mainly including multi-parameter fusion of driver multi-degree fatigue detection, MobileNetV1-based driver distraction detection, and YOLOv3-based in-vehicle driver-occupant prohibited behavior detection, considering both driver and occupant states. Experiments show that the driver-occupant state monitoring system can effectively identify all three hazardous states and has high environmental robustness, with detection accuracy of 92.18% in daytime and 88.68% in nighttime.