A Pointer Meter Reading Method Based on Image Enhancement and Deep Learning

To achieve accurate readings of pointer instruments in complex scenes of substations and promote the development of substation intelligence, this paper proposes a pointer instrument reading method based on image enhancement and deep learning. Firstly, an improved CLAHE image algorithm is introduced to enhance image enhancement details, and then, based on the YOLOv8 network, the dial area and range digital area in the instrument image are detected, and the instrument types are classified. Secondly, a perspective transformation method based on range digital information is proposed to correct the dial image for the phenomenon of instrument tilt and rotation. Finally, the semantic segmentation technology is used to extract pointer pixels in the dial area, and the pointer deflection angle is calculated in polar coordinates to obtain the instrument reading. Finally, to verify the effectiveness of this method, experiments are conducted on instrument images in general scenes, scenes with varying exposures, and scenes with varying dial tilts. The experimental results show that the proposed method has an average citation error of 0.37% and an average recognition time of 0.17 seconds. It has good adaptability in complex environments, such as lighting effects and instrument tilting, and significantly improves the accuracy and practicality of pointer-type instrument reading recognition.