Integrating Computer Vision and Geospatial Analysis to Understand Group Behavior Dynamics in Urban Parks

Urban parks function as key components of pedestrian infrastructure and everyday mobility systems. This study presents a planning-oriented analytical framework designed to integrate spatial and temporal behavioral data derived from computer vision and geospatial analysis into the planning and design process. Its novelty lies in their integration into a unified workflow that transforms municipal CCTV footage into a spatially and temporally structured dataset, linking pedestrian group-size categories to mapped park activity zones. Two neighborhood parks were selected as case studies, both featuring similar elements. Despite their common characteristics, they differ in their spatial configuration: one is linear, while the other is trapezoidal. The study provides insights into the relationship between park design and group-size usage patterns across group sizes, including individuals, pairs, small groups (3–6), and large groups (7+). The findings highlight that pedestrian group behavior varies in relation to the scale of individual spatial features and the spatial configuration and integration of park design elements. Differences in park size and spatial configuration were accompanied by different group-use patterns: the larger and trapezoidal Remez Garden showed higher levels of individual use, whereas larger-group use was more prominent in the smaller, narrower, and linear Katznelson Garden. This approach provides proof-of-concept for repurposing city camera infrastructure as a scalable tool for generating structured spatial and temporal datasets to support pedestrian movement research and evidence-based urban planning. The findings highlight the importance of design strategies that accommodate a range of group activities, thereby enhancing the functionality and inclusivity of pedestrian movement within urban open spaces.