Spatial and Temporal Analysis of Planet Scale Vehicular Imagery Data

Vehicular traffic congestion is becoming a major problem in metropolitan cities throughout the world. Looking into the future, this becomes particularly more challenging with the emergent nature combining population explosion, number of vehicles and the organic growth of cities' infrastructure. In order to study this problem, we need the traffic data and cities' physical infrastructure and the application of robust data mining and knowledge discovery techniques on this data to identify potential bottlenecks. In this work, we propose a novel method of collecting city-wide traffic information from online vehicular traffic camera. Our resulting dataset is a several months collection of vehicular mobility traces captured from 2709 traffic web cams in 10 different cities across the world, with 7.5 Terabytes of data with 125 million vehicular images. We also collect driving distance and time between geo-coordinate pairs of street intersections for these cities. We apply spatio-temporal data mining techniques to profile these global cities and reason about their geographical backbone and provide an insight into their vehicular traffic density distribution. Our results show that: (i) High correlation between driving time and distance indicate congestion-free traffic, (ii) Traffic follow certain patterns that are stable for a long time (42 days). (iii) Traffic Congestion show high Correlation (80%) for 1-2 hour lag then decrease significantly to 25-30% for four hours lag. We believe our study help to shed light on causes of contention in the present day traffic-jams and provide an insight into the planning and development of future cities and resolution to traffic congestion.