贝叶斯优化模糊C均值的城市交通状态判别方法  

Urban Traffic State Identification Based on BO-FCM

作　　者：于靖宇 魏海平[1] 郭宏伟 蔡亚峰 李润 YU Jingyu;WEI Haiping;GUO Hongwei;CAI Yafeng;LI Run(Information Engineering University,Zhengzhou 450001,China;Tianjin University,Tianjin 350000,China)

机构地区：[1]信息工程大学,河南郑州450001 [2]天津大学,天津350000

出　　处：《测绘科学技术学报》2021年第6期653-658,共6页Journal of Geomatics Science and Technology

摘　　要：准确的城市交通状态判别是实现城市智能交通诱导、管理、控制的基础和前提,对出行推荐和城市规划等具有重大的参考价值。通过分析浮动车GPS点位数据具备的位置、速度、方向等信息,可以实时获取交通参数,反映交通运行状态。针对交通状态的模糊性、不确定性等特性,以路段行驶速度和速度方差为指标,提出一种贝叶斯优化模糊C均值的聚类算法(BO-FCM)。BO-FCM用贝叶斯算法对FCM算法的初始化参数进行优化,避免FCM陷入局部最优解而导致聚类无法收敛到最优结果,降低交通状态判别的准确度。以深圳市主干道的实测数据为例,进行BO-FCM城市道路交通状态判别算法的实验分析,结果表明,BO-FCM算法较其他FCM算法,鲁棒性更高,聚类结果更准确。Accurate identification of urban traffic status is the basis and prerequisite for urban intelligent traffic guidance, management and control, which has momentous reference value for travel recommendation and urban planning. By analyzing the information of such the position, speed, direction of floating cars’ GPS data, real-time traffic parameters can be obtained to reflect the road traffic performance. In accordance with the characteristics of fuzziness and uncertainty of traffic state, a Bayesian Optimization-based Fuzzy C-Means clustering algorithm(BO-FCM) of urban traffic state division method is proposed, in which took the speed and speed variance of road sections are taken as indexes. In BO-FCM, the initialization parameters of FCM are optimized with the BO to avoid FCM’s slow convergence and low accuracy of the clustering results caused by falling local minimum. Finally, taking the GPS data of floating cars of Shenzhen main roads as an example, the experimental analysis of urban road traffic state based on BO-FCM shows that the BO-FCM has stronger robustness and better clustering effect than other improved FCM.

关 键 词：浮动车GPS数据 交通状态判别 模糊C均值 贝叶斯优化 高斯过程上置信界 

分 类 号：P208[天文地球—地图制图学与地理信息工程]