An Intelligent Self-Service Vending System for Smart Retail

Area of Science:

• Computer Vision
• Artificial Intelligence
• Retail Technology

Background:

• Traditional retail processes for non-barcode items are manual, inefficient, and error-prone.
• There is a need for automated solutions to improve accuracy and reduce operational costs.

Purpose of the Study:

• To develop an intelligent self-service vending system for non-barcode items.
• To integrate real-time product detection, weighing, and online settlement.
• To enhance the retail experience through automation and accuracy.

Main Methods:

• Implementation of a single-camera embedded system for real-time, multi-product detection without labels.
• Application of deep learning for product identification and real-time multi-object detection.
• Development of a multi-device data management platform with seamless integration with payment systems (WeChat, Alipay) and databases.

Main Results:

• Achieved a correlation coefficient (R²) of 0.99 between measured and actual weights.
• Demonstrated a non-barcode item prediction accuracy of 93.73%.
• Successfully validated system performance in diverse real-world sales scenarios.

Conclusions:

• The proposed intelligent vending system effectively automates the weighing, identification, and settlement of non-barcode items.
• The system offers a highly accurate and efficient solution for modern retail challenges.
• The technology shows significant potential for improving customer self-service and operational efficiency in various sales environments.