Semantic behavioral sequences from UAV video: Decoding the effects of the micro-scale built environment on human behavior in neighborhood centers

Understanding how neighborhood centers support behavioral processes in everyday routines is important for clarifying their contribution to sustainable neighborhood development. Such behavior is diverse, context-dependent, and often unfolds as continuous processes organized around goals and needs, calling for a person-centered behavior mapping paradigm. Yet recording such processes at the micro-scale without losing behavioral semantics, and transforming them into comparable and analyzable evidence, remains challenging. To address these challenges, this study uses semantic behavioral sequences as an analyzable representation of behavioral processes, preserving behavioral semantics while enabling standardized cross-site analysis. Using unmanned aerial vehicle footage from 43 neighborhood centers, the study integrates computer vision to infer behavior across human–environment and social interaction dimensions, sequence analysis to identify typical behavioral patterns, and manual annotation to quantify environmental features. Statistical models then test how these features and their interactions relate to site-level pattern proportions. Three behavioral patterns are identified: task-oriented, unplanned interaction, and active stopping. The models reveal distinct built-environment associations across patterns, along with significant interaction effects between function and design. This study proposes a semantic behavioral sequence framework that makes behavioral processes computable and offers a new methodological pathway for environment–behavior research. The findings provide actionable insights for targeted spatial interventions and precision design in neighborhood centers.