Abstract: With the wide application of license plate matching technology, travel times can be measured with high accuracy. However, there exist tremendous outliers in the raw data due to many reasons. The raw data can be utilized unless the outliers are filtered out. First, this paper investigates the causes of outliers’ production. Then, an overview of existing methods has been conducted to analyze potential problems with those methods. A new method, based on transport information granular computing, is proposed. An urban arterial of Delft, the Netherlands, has been selected as the test site. A comparison between the new method and existing methods has been conducted with the empirical data. The results show that the new method outperforms the existing ones.

Key words: transport information granular computing, license plate matching, travel time, outlier detection

摘要： 车牌照匹配技术的广泛应用为行程时间样本数据的精确获取提供了一种可行的方式。但是，由于各种客观原因，采集到的原始数据夹杂着大量的噪声数据，只有剔除了这些噪声数据，行程时间样本数据才能更好地应用于行程时间分析与服务。本文对产生噪声数据的原因进行了详细的分析，回顾了现有的噪声剔除方法，并分析其可能存在的问题；基于交通信息提取计算模型，提出了一种噪声剔除的新方法。以荷兰代尔夫特的一条城市道路为研究对象，基于实际数据对新方法和现有方法进行了对比分析。结果表明，新方法能够更好地剔除噪声数据。

关键词: 交通信息提取计算, 车牌照比对, 行程时间, 噪声剔除