基于小波变换的交通流数据集成灵敏度分析

Sensitive Analysis of ITS Data Aggregation via Wavelet Transform-based Algorithm

道路上的交通流数据通常是以20-30秒间隔采集并传递到交通管理中心的，为了使这些数据能够广泛地运用于各种交通用途，必须采用合理高效的集成方法对其进行处理。传统的数据集成方法基于数理统计的理论，该方法不能有效地去除数据中的噪音。近年来迅速发展的基于小波变换的交通流数据集成方法克服了这一缺点，通过对交通流数据的频率特性进行详细分析能够使集成更加高效，准确满足不同的交通用途对集成的需求。但该方法尚未作进一步的灵敏度分析，本文通过对数据集成中小波函数和数据样本量这两个重要参数的分析，从理论的角度探讨了这两个参数变化对数据集成的影响。然后以北京市三环上的交通流数据为例，计算得到了数据集成建议采用的小波族和小波阶数，以及合理的最小样本量。该结论对数据集成在将来的实际应用具有指导意义。

Traffic flow data are usually collected and transmitted to the traffic management center at an interval of 20-30 seconds. To make these data more widely used for various transportation purposes, it is necessary to process the data with a reasonable and efficient aggregation method. Conventional aggregation techniques concentrated on the statistical theory and cannot eliminate the undesired information （e.g., error or noise）. The recent development of the wavelet based aggregation technique has overcome this shortcoming, in which the frequency characteristics of traffic flow data are analyzed in details enabling the aggregation more efficient and accurate, so it can meet the demands of various transportation purposes on the aggregation. However, the sensitivity analysis on this method has never been attempted. This paper is intended to identify analytically the effects of two important parameters on the aggregation, wavelet type and sample size. Then, the paper uses the traffic flow data from the third-ring express road in Beijing as an example, to acquire the recommended wavelet family, wavelet order and the minimum sample size for the purpose of aggregation. This conclusion has guiding significance to the real applications of the data aggregation in the future.