顾及人-路交互关系的多特征协同城市道路网优化选取方法

Road network selection is an important means to realize multi-scale representation of the road network data. A number of road network selection methods have been proposed and they are mainly based on the geometry, topology, semantic information and the structural features of the road network. However, the importance level of the road for human traffic activities (i.e. the interactive relationship between human and roads) which is an important context or factor for the process of road network selection has not been fully considered by existing methods. This leads a certain mismatch between the selection results and people's actual needs. To this end, this project explores the cognition and evaluation of urban roads from the perspective of human-road relationship and develops urban road network optimization selection models and methods that take into account the human-road relationship. This project mainly includes three related issues: (1) to develop a method to integrate the multi-modal spatiotemporal data related to human travel and the road network data; (2) to study the expression of human-road interactive relationship, and to explore the road cognition knowledge contained in human travel data; (3) to construct a multi-feature collaborative road network optimization selection method by taking into account the human-road relationship and the geometry, topology and semantic features of roads, and to apply the proposed models and methods for the selective omission of urban complex road features. The research results of this project will enrich the theories methods of cartography and map generalization, promoting the research of map generalization in the big data era. On the other hand, the research results of this project will provide multi-scale road network data for applications such as urban intelligent transportation and urban planning.

道路网综合选取是实现道路网数据多尺度表达的重要手段。现有方法主要依据道路本身的几何、拓扑、语义及结构特征进行道路网选取，未充分考虑道路在交通网络中发挥的实际价值，忽略了人与道路的交互关系这一影响道路网综合选取的重要上下文信息，导致选取结果与人们的实际需求存在不一致。为此，本项目从人-路关系角度出发，针对城市道路网综合选取的复杂场景，深入研究顾及人-路交互关系的城市道路网优化选取方法，主要包括：（1）多模态人类出行时空数据与道路网的一致性整合；（2）人-路交互关系特征的定量化表达，以挖掘人类出行时空大数据中蕴含的道路重要性认知知识；（3）融合人-路交互关系特征与道路的几何、拓扑、语义特征，构建多特征协同的城市道路网优化选取模型与方法。本项目研究成果一方面将丰富地图综合的理论方法，提升道路网综合选取质量，另一方面亦可为智慧交通系统、城市规划等领域应用提供多尺度道路网数据服务。

本项目针对现有道路选取方法主要忽略了人与道路的交互关系这一对道路网综合选取具有重要影响的上下文信息，导致道路网选取结果与人们的实际需求不一致的问题，从人地关系视角出发，构建了顾及人-路交互关系的多特征协同城市道路网优化选取模型与方法。取得主要成果如下：①提出了基于人类出行微扰响应的城市道路相关性分析方法，考虑到交通流的时空非平稳性会导致路网相关性的动态不确定性，通过人类出行微扰响应过程模拟来揭示城市道路网中不同道路之间的时空相关性；②建立了顾及人类出行特征的道路重要性特征度量指标，提出了出行路径选择概率、区域吸引力等度量方法来刻画道路的功能语义特征，丰富了当前道路重要性评价指标体系；③构建了几何-拓扑-功能语义多特征协同优化的路网地图综合模型，将道路网选取操作建模为图中结点的二分类问题（选取vs不选取），通过图卷积神经网络模型学习不同特征的耦合机制，克服了人为设置特征权重的影响；④对道路网综合制图中的道路网更新问题进行拓展研究，提出了一种基于轨迹数据的道路网增量更新方法，通过自适应轨迹聚类，将复杂路网分解为简单路段，解决现有方法对复杂路网整体结构提取与更新存在的拓扑丢失与几何错误等问题。.项目研究成果在国内外权威期刊发表研究论文6篇，其中SCI/SSCI检索论文4篇，中文EI论文2篇，申请国家发明专利3项（已授权2项），研究了人-路交互关系特征以及道路的几何、拓扑与语义特征，从多特征协同的智能优化视角来解决复杂的城市道路网综合选取问题，构建了融合多维特征和深度图卷积网络模型的道路网地图选取方法，克服了道路网选取模型中参数阈值设置的困难，有效提升了道路网选取模型与方法的适用性与智能化水平。