Vision-Augmented Smart Cashless Cart System: A Multi-Sensor Approach to Ticket-Switching Detection in IoT-Driven Real-Time Billing
Abstract
The Smart Cashless Cart System (SCCS) embeds RFID/barcode identification, load-cell weight verification, cloud-based pricing, and multi-modal contactless payment within a shopping cart to eliminate the staffed checkout queue. A 90-day live pilot demonstrated a 97.3% scan detection rate, sub-300 ms cloud latency, a 78% reduction in checkout time, and a 75% drop in fraud incidents. However, weight-based verification has a structural blind spot: when a customer scans a low-cost item's barcode but places a visually similar, higher-cost item of equal mass into the cart a retail fraud pattern known as ticket-switching no anomaly is raised since the measured weight delta matches the expected delta exactly. This research work extends SCCS with an overhead, scan-triggered camera and a Visual Barcode Reconciliation Engine that compares the scanned SKU identity against an on-device MobileNetV3 classification of the item actually placed in the cart. The camera is proposed as a third, complementary sensing layer alongside not a replacement for the existing RFID/barcode and load-cell subsystems, since each layer covers a distinct attack surface. We elaborate on system architecture, sensor fusion logic, security threats, challenges, and limitations grounded in the 2020-2024 retail-fraud and IoT-security literature, and present literature-grounded performance projections pending field validation.