Research on Single-Pointer Meter Dial Recognition Method Based on YOLOv8

To address the critical requirement of precise reading recognition for single-pointer meters in industrial inspection scenarios, this study proposes an innovative multi-modal fusion architecture named YOLOv8-U2Net. The methodology achieves systematic breakthroughs through a cascaded deep detection framework: Initially, an enhanced YOLOv8 network enables rapid meter localization and dial semantic segmentation, followed by U²-Net based refinement for fine-grained pointer feature extraction, thereby establishing a multidimensional feature fusion mechanism. A novel dynamic dual-threshold interpretation mechanism based on feature fusion is innovatively designed to significantly enhance measurement reliability. Experimental results demonstrate that the proposed YOLOv8-U2Net framework exhibits superior performance advantages, achieving 93.3% accuracy in meter localization and maintaining average error margins within 0.02. Field validation confirms its capability to satisfy industrial-grade high-precision inspection requirements.