NSF Abstract

City centers nationwide are rapidly evolving from a traditional downtown into a vibrant "Central Activity District." Yet, city leaders, small businesses, and residents lack timely, trustworthy data on how people move through and experience these public spaces. This pilot project will convert existing security cameras into privacy-preserving "urban intelligence sensors." Instead of transmitting recognizable faces, each camera's video is processed on a small computer installed on-site, where people are represented only as anonymous motion heatmaps. The resulting insights, crowd-flow patterns, congested walkways, accessibility barriers, and potential safety risks will be shared in real-time with planners, shop owners, event venues, and the public through dashboards and mobile alerts. A strong partnership among UNC Charlotte, Charlotte Center City Partners, Central Piedmont Community College, Law Enforcement, and a network of local businesses ensures the work is co-designed with the very communities it serves. By demonstrating that advanced AI can be deployed ethically, transparently, and with measurable civic benefit, the pilot offers a replicable model for safer, more walkable downtowns nationwide.

The pilot will make fundamental Artificial Intelligence and Computer Vision innovations that transform raw pixels into high-dimensional thermal "motion tokens," enabling accurate detection of trajectories, dwell times, queue lengths, and anomalous behaviors without storing personally identifiable information. An online active-learning framework, reinforced by feedback from security staff and site managers, continuously adapts detection thresholds to the unique rhythms of college campuses, art venues, entertainment arenas, retail corridors, and nightlife districts. Real-time statistics are fused into spatial heatmaps, crowd-density graphs, and occupancy forecasts, which are served to stakeholders via a web dashboard and mobile app. Statistical modules aggregate foot traffic trends to guide lighting schedules, signage placement, staffing decisions, and public safety interventions. Research outcomes will include validated algorithms, open latent-motion datasets, and evidence-based policy recommendations, advancing urban informatics, responsible AI, and edge computing while establishing a scalable blueprint for AI-assisted city planning across the United States.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Award Abstract #2527312