A mixed urban traffic road-users classification based on automated video data analysis

Abstract

This article describes a novel approach for the classification of modes of travel in mixed traffic intersections based on video data. The classification is semi-supervised where the selected classification features included road-user movements’ characteristics such as speed, gait and cadence frequencies (for pedestrians and cyclists respectively) in addition to the estimate of the road-user occupied area. A modified version of the semi-supervised spectral clustering method is adapted where the selected labeled features identify possible relations; thereby enforcing certain constraints between features. Two case studies are demonstrated with video data collected at a roundabout in Vancouver, Canada and a U-turn crossover in Cairo, Egypt. Road-users were first detected and tracked using object recognition methods. The classification algorithm was then applied on the extracted objects trajectories to identify the corresponding modes of travel. Experiments were conducted on the two case studies along with a comparison to other related classification methods. A sensitivity analysis was undertaken to assess the impact of the constraints selection on the effectiveness of the method. A performance analysis demonstrated the robustness of the proposed classification method with an accuracy of higher than 87 percent achieved for both datasets. The experimental results showed that the method also outperformed other related classification methods. This research contributes to the literature of automated data collection and analysis of non-motorized traffic.