冰雪环境下基于CNN-BiGRU-MHA的汽车异常驾驶行为识别

打开文本图片集

中图分类号：U268.6 文献标志码：A

本文引用格式：，.冰雪环境下基于CNN-BiGRU-MHA的汽车异常驾驶行为识别[J].华东交通大学学报，2025，42（6）：91-100.

Anomalous Driving Behavior Recognition of Vehicles Based on CNN-BiGRU-MHAinIceand SnowEnvironments

PeiYulong，FanYichen

（SchoolofCivilEngineeringandTansportation，NortheastForestryUniversity，Harbin5oo，China）

Abstract：To enhance themonitoringand detectionofabnormaldriving behaviorof vehicles in snowand ice conditions，this paper proposes a data-driven method for identifying abnormal driving behaviors by integrating multichannel CNN-BiGRU with MHA. Abnormal driving data are obtained by LAIF model， combined with driving characteristics and data features under iceand snow environments，abnormal driving behavior indicators are constructed tocharacterize 6kindsofabnormal driving behavior，namelyrapid acceleration，rapid deceleration，rapid turning，rapid lane change，serpentine driving and skidding，and the ADASYN is introduced.The model proposed in this paper is compared and analysed with other models.The CNN-BiGRU-MHA detection model has an overall accuracy of 96.34% ，which is better than other detection models indicating that the model can effectively detect the abnormal driving behavior of cars in ice and snow environments，and provides a theoretical basis for early Warning of abnormal driving behavior.

KeyWords： intelligent transportation;abnormal drivingbehavior recognition;multi-head atention mechanism; multi-label classification; ice and snow environments

Citationformat： PEIYL，FANYC.Anomalous driving behavior recognition of vehicles based on CNN-BiGRUMHA in ice and snow environments[J]. Journal of East China Jiaotong University，2025，42（6）： 91-100.

汽车异常驾驶行为是导致交通事故发生的重要原因之一，准确识别异常驾驶行为可以降低交通事故发生率，提升交通安全2。（剩余13137字）