一种基于数字图像处理的粗指针仪表读数识别方法

A Method of Recognizing the Thick Pointer Instrument’s Numerical Reading Based on Digital Image Processing

摘要: 变电站中变压器上的油位表盘是监测变压器运行状态的关键设备，利用数字图像识别技术是准确获取油位表盘指针读数的有效途径之一。针对基于图像处理的油位表盘读数系统中存在的精度不够、可靠性低和实时性不好的问题，根据油位表指针的图像特征，提出一种基于连通区域面积大小先验知识的图像分割算法。并在该算法基础上实现变压器油位表盘读数的方法。由Matlab仿真实验表明，该方法能较好地从图像背景中分割油位表盘指针图像，为准确判定表盘指针读数奠定了基础，也为指针仪表读数识别提供了一种实用方法。

Abstract: The oil-level dial of transformer is the key device to detect the state of transformers in the transformer substation, and the digital image recognition technology is an effective way to accurately obtain the oil-level dial’s numerical reading. As to the inaccuracy, low reliability and bad real-time performance of the traditional image-based dial reading systems, according to the image features of oil-level dial, a segmentation algorithm based on the future knowledge of the connected region’s area is proposed. And based on this algorithm an oil-level dial reading method is implemented. The experimental results of Matlab simulation show that our method is better to segment the pointer from the background of the oil-level dial images, laying a foundation to accurately judge the reading of the dial’s pointer, furthermore it provides a practical way to recognize the oil-level dial’s numerical reading.