A Pointer Meter Reading Algorithm based on YOLOv8 and Digital Recognition

Abstract In the petroleum production monitoring field, the pointer meters are widely used. Recently, instead of manual inspection, using inspection robots to read pointer meters has become a new trend. In this paper, combined with the characteristics of inspection robots, based on deep learning and image processing techniques, an image-based pointer meter reading algorithm has been proposed. Firstly, the YOLOv8 detection model can be used to extract pointer meters from complex scenes; then, the YOLOv8 segmentation model can extract the pointers and scales of the pointer meters to achieve relative values; finally, the YOLOv8 detection model extract the range, the numerical recognition method can obtain the range value. The experiments shows that the prosed algorithm has real-time performance, good robustness and accuracy, which can be widely applied in the monitoring area of petroleum production.