基于数据可视化的区域交通状态特征评价方法

引入交通拥堵时空累积指标对区域交通运行状态进行判别与定量分析, 建立拥堵源与拥堵评价点之间的函数关系以构建可视化模型, 利用梯度方向直方图与主成分分析法对交通运行状态数据进行特征提取, 并利用高斯混合聚类方法对特征数据进行聚类, 划分区域交通拥堵的空间分布模式。选择了广州23 478辆出租车, 得到了509 376个数据样本, 并对交通路网拥堵模式进行识别和划分, 分析和评价了城市交通拥堵分布特性。试验结果表明: 3个样本的拥堵强度平均值分别为0.558、0.559、0.559, 交通拥堵聚集度指标分别为3.518、3.121、2.800, 3个样本的整体路网拥堵强度相同, 但是空间拥堵分布有较大差异; 广州在拥堵等级为6时的3种拥堵模式的主要聚集区域数分别为2、4、3个, 符合实际调查结果。可见, 本文提出的方法能更科学直观地描述区域交通拥堵程度和分布情况以及空间层面上的城市区域交通运行规律。

Regional traffic state evaluation method based on data visualization

The spatio-temporal accumulation index of traffic congestion was introduced to identify and quantitatively analyze regional traffic operation state.The functional relationship between congestion source and congestion evaluation point was built to construct visual model.The histogram of oriented gradient and the main component analysis method were used to carry out the characteristics extraction of traffic operation state data.Gaussian mixture clustering method was used to cluster characteristic data and devide the spatial distribution model of regional traffic congestion. 23 478 taxis in Guangzhou were chosen and 509 376 data samples were obtained.The congestion patterns of traffic road network were recognized and partitioned, and the distribution characteristics of urban traffic congestion were analyzed and evaluated.Test result shows that the average congestion intensities for three traffic samples are 0.558, 0.559 and 0.559 respectively, the traffic congestion aggregation indexes are 3.518, 3.121 and 2.800 respectively, so total road network congestion intensities for three samples are same, but there are big differences on spatial congestion distributions.When the congestion level is 6 in Guangzhou, the numbers of main aggregation regions for three congestion models are 2, 4 and 3 respectively, which accords with the practical survey result.The method proposed in this paper can depict the degree and distribution of regional traffic congestion, and describe the urban regional traffic operationalpatterns in spatial dimension more scientifically and directly.