Towards Instrumented Fingerprinting of Urban Traffic: A Novel Methodology using Distributed Mobile Point-of-View Cameras

Amidst the replication crisis, it is increasingly clear that we need to understand contextual factors that drive participant behavior, because those factors influence the applicability of study findings more broadly. For AutoUI, as we conduct interaction studies involving drivers, pedestrians, and other traffic participants, it is useful to characterize the traffic contexts that human participants are familiar with, because their prior experiences with traffic are likely to influence their behaviors within the context of a controlled study. To address this, we propose a new method, 'POV (point-of-view) camera-driven urban fingerprinting,' which can be used to characterize differentiating features of urban environments. We introduce two approaches, Small-Scale Custom Instrumentation, and Large-Scale Collection and Aggregation, and show how they can be used to acquire a broad picture on the characteristics of any city. One key benefit of POV camera-based data collection is that it better captures the experiential aspects of traffic and road scenes than methods such as satellite imaging. This work is the first to formalize a specification for this data collection method by describing existing work and outlining standards to serve as a baseline for downstream research. This work posits the medium of crowd-sourced POVcam as a new and useful tool for transportation/ automotive interaction studies and infrastructure analysis. Subsequently, we provide future researchers with guidelines for characterizing urban traffic in cities for interaction study design.