Development of an automated self-billing counter with barcode scanning and image processing on raspberry PI

Khairul Saleh, Nurul Hazreen (2025) Development of an automated self-billing counter with barcode scanning and image processing on raspberry PI. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full Text) Development of an automated self-billing counter with barcode scanning and image processing on raspberry PI.pdf - Submitted Version Download (2MB)

Abstract

Nowadays, self-billing systems have gained popularity in retail environments. This is because it offers convenience and efficiency to customers. However, there are a few challenges with the self-billing systems. This includes theft, inaccurate item identification, and slow processing times. Therefore, in this work, an automated self-billing counter that integrates image processing and a barcode reading technology on Raspberry Pi has been developed. The primary focus of this work is to enhance the self-checkout system in businesses such as small supermarkets or large hypermarkets while reducing the required number of cashier machines. The system leverages image recognition and processing techniques to detect the products before they are weighed. Teachable Machine, TensorFlow, and OpenCV were employed for image classification, while the hx711 library was used for weight measurement to verify the type of item. The performance of the developed automatic self-billing system is evaluated by considering nine different products where five products through barcode scanning and the remaining four products are through image processing. Additionally, customer purchase history and total bills are logged into Google Drive for management purposes, and receipts are sent to customers via WhatsApp using Twilio. By implementing these technologies in the automatic self-billing system, the customer will gain a good quality of services such as an efficient and seamless purchasing process. Several situations were investigated, including product detection under various lighting circumstances, item recognition from different angles, and weight measures to ensure data consistency. The results show the accuracy for each product class exceeds 70%, indicating the possibility of a more reliable and secure self-billing system.

Actions (login required)

View Item