Research on Subway Driver Behavior Recognition Method Based on Multi-Scale Feature Fusion CNN

This study proposes a computer vision-based approach for analyzing driver behavior in rail transit scenarios. The proposed method identifies irregular operating behaviors to ensure vehicle safety, addressing the inaccuracy of traditional convolutional neural networks in driver behavior analysis. This paper proposes an efficient multi-scale feature fusion network (BiFPN) model combined with dynamic space perception detection (Dynamic Head). Based on the MobileNetV4 convolutional neural network architecture, this paper significantly reduces the number of parameters and calculations while extracting features. Secondly, it also quotes the new architecture component universal inverted bottleneck to solve the inaccuracy of identifying behavior classification during driving and conducts comparative experiments with the current mainstream recognition methods to verify the advantages of this method in improving feature extraction accuracy.