Research on fatigue driving detection network based on improved YOLO11 in infrared monitoring

As deep learning prevails in intelligent transportation systems, fatigue driving detection in infrared monitoring, leveraging object detection techniques, has garnered widespread attention in both academic circles and real-world applications. However, it remains challenging to accurately capture features associated with driver fatigue across different scales due to the complexity of driving scenarios. To enhance feature extraction and fusion capabilities, this paper proposes a novel fatigue driving detection network, namely YOLO11-FDD. Firstly, the Adaptive Kernel Convolution (AKConv) is incorporated into the network to improve its performance in deal with driver states. Secondly, the spatial pyramid pooling module is optimized to better integrate features from different levels, thereby improving detection capabilities. Finally, experiments validate that the proposed network achieves satisfactory results in both robustness and accuracy. In practical applications, YOLO11-FDD can be deployed in on-board infrared monitoring systems of various vehicles (such as passenger cars, trucks, and buses) to realize real-time and high-precision fatigue driving warning.