AN INTELLIGENT MOBILE PLATFORM TO BRIDGE COMMUNITY INCIDENT REPORTING AND POLICE RESPONSE USING AI-BASED CLASSIFICATION AND REAL-TIME GEOSPATIAL HEATMAPPING
Abstract
Community public safety has traditionally relied on a one-way pipeline — a resident calls, and a dispatcher and officer respond with no shared visibility into the surrounding situation, a gap that especially discourages reporting in communities with lower trust in police. GuardLink addresses this with a Flutter mobile application that lets a resident submit one photo and one sentence, automatically classifies the incident type with an on-device keyword-scored classifier, anchors it to a location routed toward the nearest officer, and immediately contributes it to a real-time, time-windowed risk heatmap shared by residents and officers alike. Implementation challenges centered on classifier confidence under ambiguous input, balancing location privacy against officer routing, and keeping live/time-decayed incident state consistent. Two experiments run directly against the shipped code found that classifier confidence does not distinguish genuine keyword matches from fallback guesses, and that the map’s "hot" urgency flag decays independently of the selected time window. GuardLink shows a lightweight, auditable pipeline can meaningfully improve two-way community-police collaboration today.