Analysis of Automated Digital Multimeter Readings Using LabVIEW OCR and YOLOv8s-Based Object Detection

Automation of measurement data acquisition is particularly important when digital instruments do not provide a direct communication interface or when long-term measurements require high repeatability and reduced operator involvement. This paper presents a comparative study of two image-based optical reading methods to acquire indications from a digital multimeter. The reference signal was generated using a Fluke calibrator, while the multimeter display was recorded with an industrial camera. The acquired images were processed independently using two different recognition strategies. The first approach was implemented in the Python environment using a You Only Look Once version 8 small (YOLOv8s) deep learning object detection model trained to classify individual display characters. The second approach was implemented in the Laboratory Virtual Instrument Engineering Workbench (LabVIEW) using a template-based optical character recognition (OCR) method prepared in NI Vision Assistant. Acquisition series were performed for different voltage ranges, with 100 samples acquired for each pipeline and measurement case. All image-acquisition experiments were conducted under controlled laboratory conditions using a fixed camera position, a constant region of interest (ROI), and stable illumination. The detected or recognised characters were reconstructed as numerical values and compared with the reference settings of the calibrator on statistical parameters that describe the dispersion within the series and the deviation of the readings. The study evaluates the practical applicability of both approaches and demonstrates the potential of non-invasive optical reading methods for automated acquisition of indications from measurement instruments.