摘要
目前变电站仪表识别方法易受到电信号干扰,导致识别图像中存在噪声。该文结合图像识别与中值滤波方法除去图像噪声,提高仪表识别准确性。根据图像识别预处理巡检机器人采集到的仪表图像;利用颜色图像区域搜索进行目标仪表图像区域定位,依据中值滤波去除目标图像噪声;采用脉冲耦合神经网络对仪表图像数字显示盘中的数字字符实行分割和二值化处理;通过样本匹配算法匹配仪表图像样本的数字字符,实现变电站仪表数据识别。通过实验表明,基于图像识别的方法可有效识别模糊以及缺失变电站巡检机器人仪表读数,且识别准确性高。
At present,the identification method of substation instrument is easy to be interfered by electrical signal,which leads to noise in the recognition image.Therefore,combining image recognition and median filter to remove image noise,improve the accuracy of instrument recognition.Instrument image collected by inspection robot based on image recognition preprocessing.Using color image region search to locate the image region of the target instrument,and using median filter to remove the noise of the target image.Using pulse coupled neural network to segment and binarize the digital characters in the digital display panel of instrument image.The digital character of the image sample of the instrument is matched by the sample matching algorithm to realize the data recognition of the instrument in the substation.It can be seen from the experiment that the method based on image recognition can effectively identify the fuzzy and missing readings of the inspection robot instrument in substation,and the recognition accuracy is high.
作者
郑昌庭
王俊
郑克
ZHENG Changting;WANG Jun;ZHENG Ke(State Grid Zhejiang Electric Power Corporation Wenzhou Power Supply Company,Zhejiang Wenzhou 325000,China)
出处
《工业仪表与自动化装置》
2020年第5期57-61,共5页
Industrial Instrumentation & Automation
关键词
图像识别
变电站
机器人
仪表识别
脉冲耦合神经网络
image recognition
substation
robot
instrument identification
PCNN-Pulse coupled neural network
作者简介
郑昌庭(1978),男,浙江温州市人,本科,工程师,研究方向为变电运维专业管理与研究方面。
