Development of an OpenCV-based Intelligent Object Recognition, Detection and Sorting System for Retail Automation

This paper presents an end-to-end, information-driven system that links conveyor sensor events to image-based recognition for retail automation. An ultrasonic sensor and microcontroller detect items and halts the belt, triggering a smartphone to capture photos that are uploaded to a processing host with a shared event identifier and timestamp. A lightweight computer-vision pipeline (OpenCV + YOLO) extracts object labels, attributes and bounding boxes; results and raw metadata persisted as structured CSV records. By anchoring image capture to physical events, the system minimizes unnecessary inference, ensures reliable pairing of sensor and image data, and enables immediate analytics, audit trails and human-in-the-loop verification. The approach is practical on common hardware, supports real-time operations, and provides a reproducible dataset for downstream.