中文摘要：

城市道路交通状态具有空间自相关特征。某一道路交通状态的变化会对其周边道路产生影响，故把握道路交通状态的空间自相关性是提高交通规划、交通预测水平的基础。然而，城市道路交通状态又具有空间异质性，即道路交通状态的影响扩散并非各向同性，其使得道路交通状态空间自相关性的度量更为复杂，因此仅从地理空间下道路之间的邻近关系出发进行分析有失偏颇。同时，城市道路具有拓扑结构特征和几何形态特征，二者对于交通状态自相关性的影响和制约，却未引起足够重视。本文从城市道路的拓扑结构特征和几何形态特征出发，提出了一种新的交通状态空间自相关路段识别规则，即基于交通状态变化的路段空间识别规则，通过拓扑社区发现方法刻画路段在空间上的聚集特征，同时，基于Stroke跟踪的几何形态概化来描述道路交通状态变化影响的空间异质性。结果表明，利用本文提出的识别规则产生的交通状态自相关路段集合，较仪考虑地理空间邻近或拓扑结构的识别规则更为合理，更好地揭示了城市道路交通状态的空间白相关特征。

英文摘要：

Urban road traffic is spatially autocorrelated. The change of the traffic on a certain road will alter the surrounding roads＇ traffic status. Understanding the spatial autocorrelation of road traffic is essential for traffic planning and traffic prediction. However, unban road traffic is heterogeneous in spatial, which means that the traffic interactions between neighboring roads are not always isotropy. The spatial heterogeneity of urban traffic makes the measurement of spatial autocorrelation more complex, thus only uses spatial adjacency to define the traffic autocorrelated roads cannot well reveal the characteristics of spatial autocorrelation in urban road traffic. It is worth mentioning that urban roads have topological and geometric properties, which are neglected in the pre- vious research. The aim of our research is to analyze the spatial autocorrelation of urban road traffic based on the topological and geometric properties of urban roads. We first investigated the spatial clustering characteristics of urban roads using community detection algorithm, and then depicted the spatial heterogeneity of the traffic inter- action by measuring the importance of road segments with the use of the roads＇ generalized geometric forms. Based on those analyses, we proposed a novel approach to cluster together the roads whose traffic is spatially au- tocorrelated. Experiment results for the road network of Beijing indicate that the proposed approach performs better than the approaches that only consider the spatial adjacency or topological structure, which further implies that our approach can capture the spatial autocorrelation characteristics of urban road traffic more reasonably.