摘要
在工业现场中,通过移动机器人自主巡检与读表的应用越来越广泛。数显仪表的读数受到光照、对焦和场景等因素的影响,存在识别精度低的问题。针对上述问题,提出了融合感知与控制的数显仪表识别机制,首先使用图像目标检测网络识别数显仪表,标出目标区域。随后设计基于图像反馈的智能车控制策略,以此抑制模糊对焦的影响。再使用改进的自动色彩均衡(ACE)算法对基于色彩膨胀切割的数字块进行图像色彩增强。最后通过卷积循环神经网络(CRNN)进行实时仪表数字识别。通过实验验证了所设计技术的性能,证明了方法的可行性和有效性。
In the industrial field,the application of autonomous inspection and reading tables through mobile robots is more and more widely.The reading of digital display instrument is affected by factors such as illumination,focusing and scene,which has the problem of low recognition accuracy.Because of the above problems,this paper proposes a digital display instrument recognition mechanism based on fusion perception and control.Firstly,the image target detection network is used to identify the digital display instrument and mark the target area.Then an intelligent vehicle control strategy based on image feedback is designed to suppress the influence of fuzzy focus.Then the improved automatic color equalization(ACE)algorithm is used to enhance the image color of digital blocks based on color expansion cutting.Finally,real-time instrument digital recognition is acheived using convolutional recurrent neural network(CRNN).In this paper,the performance of the designed technology is verified by experiments,and the feasibility and efficiency of the method is proved.
作者
吴开田
陈丹
邱洋
徐哲壮
简荣贵
陈剑
Wu Kaitian;Chen Dan;Qiu Yang;Xu Zhezhuang;Jian Ronggui;Chen Jian(College of Electrical Engineering and Automation,Fuzhou University,Fuzhou 350000,China;Key Laboratory of Industrial Automation Control Technology and Information Processing,Education Department of Fujian Province,Fuzhou 350000,China)
出处
《信息技术与网络安全》
2021年第10期53-60,共8页
Information Technology and Network Security
基金
国家自然科学基金面上项目(61973085)。
关键词
数字识别
感知与控制
图像色彩增强
深度学习
digital recognition
perception and control
image color enhancement
deep learning
作者简介
吴开田(1994-),男,硕士研究生,主要研究方向:机器视觉、深度学习;陈丹(1977-),女,博士,副教授,主要研究方向:机器人视觉伺服控制。;通信作者:徐哲壮(1984-),男,教授,博士生导师,主要研究方向:工业物联网技术与应用。E-mail:zzxu@fzu.edu.cn。
