A Trip Building and Chaining Methodology Using Traffic Surveillance Data

Abstract

In many cities, traffic video surveillance systems have been installed at major intersections. These cameras can capture not only the traffic flow or violations but also the time, location, driving direction, color, and license plate of vehicles. This paper proposes an approach to build trips based on video surveillance data, combine the trips to form daily travel chains, and efficiently classify all travel chains into different modes. A K-means method finds clusters of different types of vehicles, and exclude profitable vehicles, which are always on the road. Four trip chaining patterns are derived from the data and used as the training set. A support vector machine (SVM) method classifies the daily trip chaining patterns. The results show that video surveillance data contains rich information on the traffic patterns and can be used in building the trips. The SVM method can classify trip chaining patterns efficiently with excellent results when processing a large amount of data.