关键词: 伴随车辆, Spark计算框架, FP-Growth算法, 随机伴随, 片伴随

Abstract:

With the extensive application of big data technology in traffic management, it has aroused the researchers’ attention to detect the companion vehicles in the massive license plate data. However, most of the current methods are inefficient at large data volumes and remain in the theoretical research stage without being combined with practical applications. This paper presents a novel approach to this application. Using Spark distributed parallel computing framework to improve the running speed, the load balancing principle is used to equalize the data, and then a companion vehicle discovery algorithm based on the improved FP-Growth is proposed. The confidence is used to post-process the results, excluding the random companion situation, improving the detection accuracy. The method is applied to Changsha Traffic Police Major Traffic Control Center System, in which massive license plate recognition data is stored in Hive database under Hadoop big data platform, visualized on Police PGIS （Police Geographic Information System）. The experiment proves the efficiency and feasibility.

Key words: companion vehicles, Spark computing framework, FP-Growth algorithm, random companion, slice companion