高速公路隧道空间有限，内外亮度差异明显，驾驶员在隧道不同位置的驾驶行为差异较大，全路段统一的预警和管控难以精准地实现对隧道不同路段的差异性管理。基于同济道路轨迹数据平台 （Tongji road trajectory sharing platform, TJRD TS），提取车辆连续微观参数，以8项指标量化驾驶员行车特性，以此解析车辆在不同位置的驾驶行为和交通安全风险差异。基于无监督学习算法，提出一种考虑行车特性的高速公路隧道路段分段方法：首先运用主成分分析法 （principal components analysis, PCA）确定表征驾驶行为和交通安全的主要特征。随后通过K-means聚类算法将主要特征沿隧道方向的分布进行划分，最终结合显著性分析验证隧道路段划分的合理性。结果表明：驾驶员在隧道不同位置的驾驶行为和交通安全性具有较大的差异；根据行车特性，运用PCA-K-means聚类将隧道路段划分为接近段、入口段、过渡段、中间段、出口段、驶离段6个部分；入口段及过渡段车速变化离散性较大，交通流不稳定；过渡段、出口段交通冲突频发，并且车辆减速比例、加速比例分别达到最高值14.89%、15.65%。研究成果揭示了隧道内车辆行车特征的演化规律，基于此对高速公路隧道进行有效分段。研究成果有利于隧道车辆主动安全控制策略的制定以及精准的车路协同管控实现。

The confined space and fluctuating brightness levels inside and outside highway tunnels result in notable disparities in driving behaviors across various sections. It's difficult to achieve differential management of various sections within tunnels due to the challenge of implementing uniform warning and control across the entire roadway. Based on the Tongji road trajectory sharing platform(TJRD TS), continuous microscopic parameters of vehicles were extracted to quantify driving characteristics using eight indicators. This approach was aimed at analyzing the differences in driving behavior and safety risks of vehicles at different tunnel locations. Based on unsupervised learning algorithms, a segmenting method was proposed for highway tunnel sections that considers driving characteristics. Firstly, principal components analysis (PCA) was employed to determine the main features representing driving behavior and traffic safety. Subsequently, the K-means clustering algorithm was utilized to divide the distribution of main features along the tunnel direction into segments. Finally, the rationality of tunnel section division was validated through significance analysis. The results show that the driving behavior and safety vary significantly at different positions within the tunnel. Based on driving characteristics, the tunnel sections are segmented into six parts using PCA-K-means clustering: approach section, entrance section, transition section, middle section, exit section, and departure section. The entrance and transition sections exhibit high variability in speed changes and unstable traffic flow, while conflict frequencies are high in the transition and exit sections, with vehicle deceleration and acceleration reaching peak values of 14.89% and 15.65%, respectively. The research results reveal the evolution pattern of vehicle driving characteristics within tunnels and facilitates effective segmentation of highway tunnels. The research results contribute to the formulation of proactive safety control strategies for tunnel vehicles and the realization of precise vehicle-road cooperative control.