Vision transformer-based incident detection algorithm for freeway service areas

Faced with the loss of incident detection models for freeway service areas, a vision transformer-based model was proposed in this paper. And the proposed model was used in different scenes, such as queueing in gas station and charging station, fire, heavy fog, garbage overflow, smoke and make phone calls in gas station and illegal parking. An experiment was conducted in JiangBei Service area for 30 days, and the accuracy and recall rate are utilized to evaluate the proposed model and other 3 models. And the proposed model in this paper outperformed 3 other models in most scenes, which indicates that the trained model has strong ability to generalize abnormal event detection.