摘要
针对机械式水表读数的现有图像识别方法识别成功率低、效率低的问题,开发了一种动态识别指针(或梅花轮)位置的图像识别方法。对被测水表进行摄像,自动定位、跟踪末位指针或梅花轮转动圈数和角度(坐标),实时图像识别转化为水量读数,结合计量标准器自动测量的实际累积水量,实现水表示值误差自动检定。设计了一套基于图像识别的质量法机械式水表串联自动检定装置,实现6~10台同规格DN(15-25)水表同时自动检定;图像识别成功率高于99.9%,检定装置准确度等级达0.2级;与人工检定相比,检定时长从23.5~43分钟降低至11~15分钟,显著提升了水表检定工作质量和效率。
A water meter reading image recognition method based on dynamic recognition of the position of the pointer(or plum blossom wheel) is developed to address the low success rate and efficiency of existing image recognition methods for mechanical water meter readings.By capturing the measured water meter,automatically locate and track the number of turns and angles(coordinates) of the last pointer or plum blossom wheel,convert real-time image recognition into water volume readings.An automatic verification device for mechanical water meter series based on quality method and image recognition has been designed to achieve simultaneous automatic verification of 6~10 DN(15-25) water meters of the same specification.The success rate of image recognition is higher than 99.9%.Compared with manual verification,the verification duration has been reduced from 23.5~43 minutes to 11~15 minutes.
出处
《工业控制计算机》
2024年第5期52-54,共3页
Industrial Control Computer
基金
广州市市场监督管理局科技项目:基于激光传感器的DN(125-200)机械式水表智能检定装置研制(2022KJ24)
基于图像识别的DN(15-25)机械式水表智能检定装置研制(2023KJ21)。
关键词
机械式水表
自动检定装置
读数
图像识别
质量法
串联
mechanical water meter
automatic verification device
reading
image recognition
quality method
series
