Automated Pedestrian Safety Analysis Using Video Data in the Context of Scramble Phase Intersections

Pedestrian data is critical for supporting safety studies and proper design of pedestrian facilities. There has been a traditional shortcoming in the transportation literature with regard to the availability of and quality of pedestrian data. A potential reason is that current data collection methods mainly rely on human observers which make them costly and prone to observer-based errors. In addition, automated data collection methods currently in practice are mainly developed to serve motorized traffic with few proven applications at similar level of functionality for pedestrians. This paper describes a novel technique for conducting road safety analysis using an automated video analysis system. The system was tested using video data collected from a scramble phase intersection in the downtown area of Oakland California. The context of scramble phase is unique in that conflict indicators have to be developed further in order to represent the severity of interaction outside crosswalks. This study demonstrated the feasibility of using computer vision techniques to automatically analyze pedestrianvehicle conflicts. The ability to analyze pedestrian-vehicle conflicts will potentially reduce the cost of conflict surveys and improve the quality and quantity of safety-related data. In addition, pedestrian walking speed data was collected using other video sequences at high accuracy. An accurate measurement of pedestrian walking speed is critical for understanding pedestrian behaviour at crosswalks and for properly designing traffic signals. To the best of the authors’ knowledge, there is little, if any, work similar to this study that exists in the technical literature in regard to the method of analysis and context of scramble phase intersection.