地理格网模型支持下的轨迹数据管理与分析框架:方法与应用

Trajectory Data Management and Analysis Framework Based on Geographical Grid Model:Method and Application

近年来,各类位置感知设备产生的轨迹数据被广泛应用于城市规划、智能交通、公共卫生、行为分析等各个领域,但是常规矢量表达方式及建立在其基础之上的分析算法的计算复杂度高,无法满足大规模轨迹数据的高时效性应用需求。针对上述问题,提出了基于地理格网模型的轨迹数据管理与分析框架,为轨迹数据挖掘的"表达-管理-分析-应用"全链条研究提供新的技术框架,主要包含地理格网模型、轨迹多尺度表达与组织、轨迹计算与分析、高性能计算技术、轨迹挖掘应用5部分。介绍了各部分的实现思路和方法,并阐述了格网模型与轨迹数据结合的优势,包括存储管理高效灵活、适合高性能计算技术和契合自动控制与智能计算需求等。以城市交通流多层级实时可视化和基于地理格网编码的相似性分析两个应用实例验证了该技术框架理论与技术方法的可靠性和有效性。

Objectives: In recent years, trajectory data generated by various location-aware devices have been widely used in a host of fields such as urban planning, intelligent transportation, public health, and behavior analysis, and have produced huge social and economic value. However, the conventional vector representation model and the analysis algorithms based on it have high computational complexity and cannot meet the high time-efficient application requirements of large-scale trajectory data. To address the above problems, we propose a trajectory data management and analysis framework based on a geographical grid model, which provides a new technical framework for the entire chain of "representation-management-analysis-application" of trajectory data mining.Methods:This framework includes five parts:(1) Construction of geographical grid model. The Earth. s surface space is subdivided level by level according to some rules(e.g., quadtree, octree), and the spatiotemporal position is represented by grid coding.(2) Multiscale coding representation and organization of trajectories. The trajectory data are mapped onto different spatial resolution grid levels according to the spatial position information and precision requirements to im-plement the spatial one-dimensional multi-scale coding representation instead of multi-dimensional vector coordinates, to reduce the difficulty of organization and management on massive trajectory data.(3) Calculation and analysis of trajectories. The concept of geographical grid calculation is proposed, and the existing complex algorithms are modified by using low complexity coding operations(including intrinsic index,multi-scale, set operations, etc.) to accelerate the process of trajectory data mining analysis.(4) High-performance computing technologies. We further utilize the discreteness of geographical grid model, and combine the current parallel and distributed high-performance computing frameworks to achieve distributed storage and concurrent computing on trajectory big data.(5) Applications of trajectory mining. Based on this framework,a series of high time-efficient applications are proposed to better serve the fields of urban management, intelligent transportation, and public safety. In addition, we found several advantages of combining geographical grid model and trajectory data, including large-scale data storage and management efficiency and flexibility, suitability for high-performance computing technology, and meeting the needs of automatic control and intelligent computing.Results: Supported by the above theoretical methods, the reliability and effectiveness of the technical framework theory and technical methods have been verified by two applications-multi-level real-time visualizing of urban traffic flow and similarity analysis based on geographical grid model. The results show that the spatiotemporal query and similarity measure under this framework are one to two order of magnitude higher than classic algorithms under the vector coordinates. It has achieved to response within seconds and real-time visualization while processing billions of trajectory data, which proves the reliability and validity of the theories and technical methods. We explain the content of the proposed trajectory data management and analysis framework from the perspectives of framework design, theoretical analysis, advantages summary, and comparative experiments. Conclusions: It has been verified that the framework has the potential for high-performance management and analysis of massive trajectory data. In the future, issues such as the tradeoffs between data volume and accuracy, complex network analysis, and multi-source element association analysis need to be further explored and studied, to build a complete trajectory data management and mining technology system in a geographical grid space.