车辆轨迹地图匹配异常成因辨识方法

Methodology of Identifying Causes of Abnormal Map Matching

摘要:

车辆轨迹地图匹配异常指轨迹被匹配到不合理的地图路段，其主要成因可归纳为物理遮挡、地图路段缺失和复杂路网。针对这一现象，提出一种数据驱动的匹配异常轨迹段快速识别和成因归类方法。首先归纳不同成因导致的地图匹配异常数据特征，并构建表征指标；其次提出车辆轨迹分段方法，区分正常和异常的地图匹配轨迹段；进而建立基于随机森林的地图匹配异常轨迹段成因分类方法；最后通过上海市地图路网和出租车营运轨迹数据验证方法的有效性。验证表明，提出的方法准确率达93.5%，可有效辨识物理遮挡、社区路段缺失和复杂路网三种成因，同时具有较好的区域可迁移性。

Abstract:

Abnormal map matching refers to that partial vehicle trajectory is matched to a road segment incorrectly， typically caused by physical occlusion， road missing， and complex road configuration. This study proposes adata-driven methodology of identifying the causes ofabnormal map matching， which supports fast identification and classification of abnormal matching results. First， we summarize the statistical features of abnormal matching results with different causes and present criteria for representation. Second， we propose a segmentation algorithm to distinguish the abnormal results from the trajectory. Third， a classification algorithm is developed based on the random forest model. Finally， the proposed methodology is validated based on the real-world data of urban taxi operations in Shanghai. The case study reveals that the algorithms are effective in identifying the abnormal matching results caused by physical occlusion， local road missing， and complex road configuration， with a precision of 93.5%. In addition， the experiments show the high spatial suitability of the proposed methodology.