Design and Evaluation of a Real-Time Barcode Detection System Using Video Analysis

Alysson Pereira do Nascimento, Ruan Gomes, Otacílio Araújo Ramos Neto

DOI: 10.14209/sbrt.2025.1571157188

Evento: XLIII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2025)

Keywords: Barcode detection computer vision Industry 4.0 applications

Abstract

This paper presents the BarcodeVision system, %and initial experimental results of the BarcodeVision system, designed for real-time barcode detection and reading in uncontrolled environments. The first version of the system implements four reading approaches, ranging from methods that use only the ZBar library to strategies with additional processing to improve detection and reading, such as multithreaded frame rotation, the use of You Only Look Once (YOLO) detection, and region-of-interest (ROI) reading. The experiments were performed using a video recorded on an automated treadmill, resized to 11 different resolutions, ranging from VGA to DCI 4K. In total, 218 frames containing readable barcodes were counted, which were used as the basis to calculate the percentage of reading and the average processing time of each method. The results showed that the method that combines ZBar with multithreaded rotation presented the highest reading percentage, evidencing the effectiveness of rotation compensation. YOLO-based methods demonstrated competitive performance, especially in scenarios with codes present, with shorter response times.

Download