iTV: Inferring Traffic Violation-Prone Locations with Vehicle Trajectories and Road Environment Data

TL;DR

This paper presents a dynamic, low-cost framework that leverages vehicle trajectory and road environment data to accurately identify locations prone to traffic violations, aiding urban enforcement strategies.

Contribution

It introduces a novel, comprehensive approach combining data normalization, behavior extraction, and pattern analysis to infer violation-prone spots from large-scale vehicle data.

Findings

Effective inference of violation-prone locations demonstrated on real-world data.

Framework provides timely and accurate insights for traffic enforcement.

Visualization system aids decision-making for urban traffic management.

Abstract

Traffic violations like illegal parking, illegal turning, and speeding have become one of the greatest challenges in urban transportation systems, bringing potential risks of traffic congestions, vehicle accidents, and parking difficulties. To maximize the utility and effectiveness of the traffic enforcement strategies aiming at reducing traffic violations, it is essential for urban authorities to infer the traffic violation-prone locations in the city. Therefore, we propose a low-cost, comprehensive, and dynamic framework to infer traffic violation-prone locations in cities based on the large-scale vehicle trajectory data and road environment data. Firstly, we normalize the trajectory data by map matching algorithms and extract key driving behaviors, i.e., turning behaviors, parking behaviors, and speeds of vehicles. Secondly, we restore spatiotemporal contexts of driving behaviors to get corresponding traffic restrictions such as no parking, no turning, and speed restrictions. After matching the traffic restrictions with driving behaviors, we get the traffic violation distribution. Finally, we extract the spatiotemporal patterns of traffic violations, and build a visualization system to showcase the inferred traffic violation-prone locations. To evaluate the effectiveness of the proposed method, we conduct extensive studies on large-scale, real-world vehicle GPS trajectories collected from two Chinese cities, respectively. Evaluation results confirm that the proposed framework infers traffic violation-prone locations effectively and efficiently, providing comprehensive decision supports for traffic enforcement strategies.