Image-based Clustering Analysis of Massive Vehicle Networking Data

Zhonghui Chen,Zhipeng Xie,Xinxin Feng

DOI: https://doi.org/10.1109/wcsp.2016.7752687

2016-01-01

Abstract:In order to make full use of the massive vehicle networking data and dig out the characteristics of the vehicle driving behaviors, the analysis of the massive vehicle networking data is indispensable. And clustering analysis is an effective way to extract the information which can guide and supervise the vehicles. In this paper, we propose a novel initial clustering center selection algorithm based on the image recognition. In addition, in order to shorter the time of the clustering algorithm, we improve the ε-Link based on the massive vehicle networking data. We refer to image-based clustering algorithm applied on matching map of the vehicle networking data and ultimately obtain the traffic hot spot map of the urban roads by the clustering analysis. We evaluate our algorithm by doing thorough simulation on the massive data of vehicles in Fuzhou city. The result shows that when compared to the traditional clustering algorithms with great randomness in selection, the image-based clustering algorithm consumes less time in the case of massive vehicle networking data. And through the analysis of the trajectory data, the traffic hot spot map can be formed and practically reflects the practical traffic flow of Fuzhou city.