基于时空大数据交通路网盲信息处理算法与实现

发布时间：2018-05-01 19:31

本文选题：盲信息处理 + 异常数据分离　；参考：《山东大学》2016年硕士论文

【摘要】：城市智能交通系统能够提高道路通行效率,是解决交通拥堵这一城市发展瓶颈的有效方法。但是智能交通系统的正常运行必须依赖于能够准确描述即时交通路网结构的电子地图。传统方法建构城市路网地图时,利用地图匹配模型将遥感数据和探测车数据进行融合,这需要漫长的时间来采集数据,而且两种来源不统一的数据之间的兼容可能会引入误差。及时更新城市交通路网地图数据的先决条件包括两个方面,即得到单条矢量道路(如道路路段)结构数据和关键点(如交通信号灯)位置数据。浮动车轨迹定位数据可以描述道路结构,另外其中包含的瞬时速度和瞬时方向角等行驶特征数据还可以描述浮动车的行驶状态。根据浮动车在某条道路上的行驶特征可以估计出该道路的某些属性参数。综上,浮动车轨迹定位数据是建构和更新交通路网的理想数据源。基于此,本文提出一种基于时空大数据交通路网盲信息处理的算法,首先提出交通数据盲信息处理理论,具体分析浮动车观测数据之间空间位置分布关系和浮动车轨迹与矢量道路的空间相遇关系；然后设计了一种按时段分离交通异常数据的方法,分离得到不存在明显异常的观测数据；通过实验提出邻域质心聚类方法并通过线性插值和弧段插值的方法得到矢量道路结构核心点；最后提出利用邻域平均速度提取道路关键点的方法,并根据矢量道路上下行对称的特征对结果进行了修正。本文方法不借助任何人工测量数据,只利用浮动车观测数据,通过盲信息处理建构矢量道路结构并提取出道路中的关键点位置(公交站点、交通信号灯位置)。本文将上述方法结合济南市出租汽车和公交车的GPS轨迹观测数据进行了实验,结果表明该方法建构的矢量道路精度较高,能够准确描述真实道路特征,提取的道路公交站点和交通信号灯位置距离真实位置的平均误差分别是6.31米和7.58米,数据更新频率是10天／次,能够快速准确实现交通路网拓扑的建构和迭代更新。
[Abstract]:Urban Intelligent Transportation system (its) can improve the traffic efficiency and is an effective method to solve the bottleneck of urban development. However, the normal operation of its depends on the electronic map which can accurately describe the structure of real-time traffic network. When the traditional method is used to construct the urban road network map, the remote sensing data and the rover data are fused by using the map matching model, which takes a long time to collect the data, and the compatibility between the two kinds of data with different sources may introduce errors. The precondition of updating the map data of urban traffic network in time includes two aspects: the structural data of single vector road (such as road sections) and the position data of key points (such as traffic lights). The track location data of the floating vehicle can describe the road structure, and the driving characteristic data such as instantaneous velocity and instantaneous direction angle can also be used to describe the driving state of the floating vehicle. According to the driving characteristics of a floating vehicle on a road, some attribute parameters of the road can be estimated. In summary, floating vehicle locus location data is an ideal data source to construct and update traffic network. Based on this, this paper presents an algorithm for blind information processing of traffic network based on space-time big data. Firstly, the theory of blind information processing of traffic data is proposed. The spatial position distribution relationship between the observation data of the floating vehicle and the spatial encounter between the track of the floating vehicle and the vector road are analyzed in detail, and then a method of separating the abnormal traffic data according to the period of time is designed. The observed data without obvious anomalies are separated and the core points of vector road structure are obtained by linear interpolation and arc-segment interpolation, and the neighborhood centroid clustering method is proposed by experiments. Finally, a method of extracting the key points of the road by using the average velocity of the neighborhood is proposed, and the results are corrected according to the symmetric feature of the vector road. In this paper, the vector road structure is constructed by blind information processing without any manual measurement data, and only the floating vehicle observation data is used. The key points in the road (bus stations, traffic lights) are extracted. In this paper, the above method is combined with the GPS track observation data of taxi and bus in Jinan. The results show that the vector road constructed by this method has high accuracy and can accurately describe the real road characteristics. The average error between the location of the bus stop and the traffic signal is 6.31 m and 7.58 m, respectively, and the frequency of data updating is 10 days / time, which can quickly and accurately realize the topology construction and iterative updating of the traffic network.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U495

【相似文献】