重型长途货车驾驶员在高速公路驾驶时易发生疲劳，威胁行车安全。为探究重型货车驾驶员在高速公路连续驾驶过程中的疲劳特征及其变化规律，特开展货车驾驶员高速公路连续驾车实车实验。使用摄录机对驾驶位进行全程监控，通过视频对驾驶员驾驶过程中观察后视镜次数、小动作次数等与疲劳程度相关驾驶行为进行统计；使用眼动仪对驾驶员驾驶过程中眨眼时间、瞳孔直径等眼部数据进行提取与计算。之后通过函数拟合、驾驶员行为特征分析、多层感知模型（MLP）构建的方法对重型货车驾驶员高速公路连续驾驶过程中的疲劳特征变化规律、疲劳特征与驾驶行为关系以及连续驾驶时间安全阈值进行研究。结果表明：货车驾驶员连续驾车时间安全阈值介于3.6~3.7 h之间；在驾驶任务过程中所产生的被动疲劳方面，依据平均眨眼时间判断驾驶员产生疲劳感在48.69%的情况下可比驾驶员出现主动疲劳干预行为提前5~15 min；重型长途货车驾驶员在连续驾驶1.5~2.0 h期间被动疲劳出现的频次减少；观察后视镜和做与驾驶无关的小动作两种行为的次数与平均眨眼时间具有相关性；使用观察后视镜次数、小动作次数与通信行为次数替代眼动指标构建多层感知器模型时，其对驾驶过程中的被动疲劳检出率可达到80.2%。该研究结果可为重型货车驾驶员疲劳的视频监测与提示以及货车行车安全性提升提供参考。

Long-distance heavy truck drivers are prone to fatigue when driving on highways, threatening road safety. To systematically investigate the fatigue characteristics and their progression of heavy truck drivers during prolonged highway driving, an on-road driving experiment was conducted for truck drivers operating continuously on highways. Cameras were mounted to continuously monitor the driver’s seat, with driving behaviors correlated to fatigue levels-such as the frequencies of rearview mirror-checking and non-driving-related movements-quantified through video analysis. An eye tracking device was utilized to acquire and process ocular metrics-including blink duration and pupil diameter-during driving sessions. Subsequent analysis employed function fitting, driving behavior characterization, and multilayer perceptron (MLP) modeling to examine the progression patterns of fatigue characteristics during prolonged highway driving, the correlation between fatigue manifestations and driving behaviors, and safety-critical time thresholds for continuous driving duration of truck drivers. The results indicate that the safe threshold for continuous driving time for truck drivers is between 3.6 and 3.7 hours. Regarding passive fatigue generated during driving tasks, detection of fatigue onset via mean blink duration occurred 5~15 minutes prior to drivers’ active fatigue countermeasures in 48.69% of observed instances. A notable reduction in passive fatigue frequency was observed during the 1.5~2.0 h interval of continuous driving. Statistical analysis revealed significant correlations between both rearview mirror-checking frequency and non-driving-related movements (NDRMs) with mean blink duration. An MLP model constructed using the frequencies of rearview mirror-checking, NDRM occurrences, and communication behaviors as predictors without ocular metrics achieved an 80.2% detection rate for passive fatigue during driving. These research outcomes provide theoretical foundations and practical references for developing video-based fatigue monitoring systems, implementing fatigue alerts, and enhancing driving safety for heavy trucks.