摘要
针对发电厂中指针式仪表数量多、周围环境复杂、灰尘污染等因素导致获得的图像含有大量噪声从而影响仪表读数识别精度的问题,提出了一种改进Canny边缘检测算法,通过采用5×5邻域计算像素梯度幅值的方法,提高了边缘检测精度,并采用自适应双阈值选择法,有效避免了传统Canny算法易出现伪边缘、边缘信息多等现象,结合Hough变换技术提取仪表指针信息,提高仪表识别精度。经实验对比验证,该算法检测效果优于传统Canny算法,仪表读数识别精度达到95%。
Due to the large number of pointer instruments,the complex environment,dust pollution and other factors in power plants,the image obtained contains a lot of noise which affects the accuracy of meter reading recognition.An improved Canny edge detection algorithm is presented.By adopting the method of 5×5 neighborhood to calculate the amplitude of pixel gradient,edge detection accuracy is improved and adaptive double threshold selection method is adopted.Compared with traditional Canny algorithm,it can avoid false edge and more edge information.The instrument pointer information is extracted by Hough transform technology,the accuracy of instrument identification is improved.Verified by comparison,the detection result of this algorithm is better than that of traditional Canny algorithm,and the accuracy of meter reading recognition is 95%.
作者
姚洋
彭道刚
王志萍
YAO Yang;PENG Daogang;WANG Zhiping(School of Automation Engineering,Shanghai University of Electric Power,Shanghai200090,China)
出处
《上海电力大学学报》
CAS
2020年第2期183-189,共7页
Journal of Shanghai University of Electric Power
基金
上海市科学技术委员会工程技术研究中心项目(14DZ2251100)。
作者简介
通信作者:彭道刚(1977-),男,博士,教授。主要研究方向为智能发电、能源互联网、电力巡检机器人等。E-mail:pengdaogang@126.com。
