Vision-Based Smart Weighing System

Abstract

Abstract The The application of automation technology in weighing systems to improve efficiency and accuracy has gained increasing attention. Limitations of conventional weighing systems, such as the inability to automatically recognize object types, limited measurement accuracy, and manual payment processes, have motivated the development of a Vision-Based Smart Weighing System. This system is designed to automatically recognize fruit types, measure weight, and perform digital payments by integrating Machine Learning, YOLO (You Only Look Once), and Convolutional Neural Network (CNN) technologies. Experimental results indicate that the YOLOv11 model achieves an average object detection accuracy of 86.4%, while fruit classification using CNN reaches an average accuracy of 91.3%. The Load Cell sensor testing produces an average absolute error of 0.033 kg with an average percentage error of 2.79%, which is acceptable for retail weighing applications. Furthermore, the integrated Midtrans payment gateway demonstrates a 100% transaction success rate across multiple payment methods, with an average transaction duration of 19 seconds and an API response time of approximately 0.0004 seconds.These results demonstrate that the proposed system successfully improves the speed, accuracy, and efficiency of automatic weighing and payment processes. Therefore, the system is suitable for implementation in supermarkets, grocery stores, and fruit shops to reduce human error and enhance customer convenience.

Keywords: Smart Weighing System, Load Cell, Visual Weight Detection, Deep Learning, System Automation and Payment.