Zone-Based Joint Camera-Kiosk Analytics for Smart Retail Insights

This study presents a novel system that tightly integrates zone-based camera analytics with kiosk interaction logs to provide rich, multidimensional insights into customer behaviour in retail environments, particularly within smart retail settings in smart cities. Our approach enables retail administrators to define arbitrary, polygonal zones on a live floorplan overlay for each camera. In the proposed pipeline, video streams are processed at 30 fps using YOLOv8 for person detection and SORT for tracking, with entry and exit events logged per zone to compute footfall, dwell time, and bounce rates. Zones are classified as kiosk zones or custom zones based on metadata assigned during zone creation in the dashboard, where each polygon is manually linked to a kiosk identifier or defined as a general area of interest. By correlating zone presence with kiosk session data via timestamp overlap, we derive a robust kiosk engagement metric. We deployed the system in a mid-size retail store over four weeks, demonstrating improvements in the granularity and actionability of behavioural analytics and revealing optimal kiosk placements and display configurations in a smart retail context.