Estimating vibrancy indicators in urban spaces using object tracking

Abstract

Evidence-based policy making (EBPM) using digital measurements of urban vibrancy is advancing in public space management and tourism-oriented community development. However, the information ob-tainable through existing image recognition technologies can be further expanded. This study proposes a method for estimating directional pedestrian flows and pedestrian dwell times from arbitrarily defined po-lygonal regions using deep learning-based object tracking.

The proposed method was evaluated using video data collected during a summer festival in the Oya district of Utsunomiya City, demonstrating that overall pedestrian flow patterns could be captured. In ad-dition, quantitative indicators of urban vibrancy, including temporal variations in the number of people within the event area and average dwell times estimated using Little's Law, were obtained. Furthermore, the analysis results were utilized as a consensus-building tool for local stakeholders, demonstrating the usefulness and practical applicability of the proposed method.