摘要
首先选用具有广角视场角的鱼眼镜头来采集含有多个指针式仪表的图像,然后对多仪表图像进行3步预处理操作,图像灰度化,消除鱼眼镜头的桶形畸变,消除由镜头平面与表盘平面不平行引起的图像形变,并将多仪表图像分割成多个单仪表图像。依次对每个单仪表图像进行读数识别,过程包括表盘提取、指针检测、读数计算。实验结果表明:在分割出的单仪表图像质量不高的情况下,仍能得到比较满意的读数结果。
Firstly,a fisheye lens with a wide-angle field of view is used to capture images containing multiple pointer meters.Then three steps of pre-processing operation are carried out for multi-meters image.Firstiy,image graying is carried out.secondly,the barrel distortion of fish-eye lens is eliminated;thirdly,the image distortion caused by the non-parallel of lens plane and dial plane is eliminated,and the multi-meters image are divided into several single-meter images.Finally,reading recognition of each single-meter image in turn,the processes include dial extraction,pointer detection,and reading calculation.The experimental results show that in the case of the segmented single-meter image quality is not high,the satisfactory reading results can still be obtained.
作者
赵健
娄震
ZHAO Jian;LOU Zhen(College of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing 210094,China)
出处
《传感器与微系统》
CSCD
2020年第2期61-64,共4页
Transducer and Microsystem Technologies
关键词
多仪表
指针式仪表
鱼眼镜头
图像畸变
读数识别
multiple meters
pointer meter
fisheye lens
image distortion
reading recognition
