一种改进的指针式仪表读数区域分割方法  

An Improved Reading Area Segmentation Method for Pointer Instrument

作　　者：李腾辉 何秋生[1] 丁思萌 LI Teng-hui;HE Qiu-sheng;DING Si-meng(Collage of Electronics and Information Engineering,Taiyuan University of Science And Technology,Taiyuan 030024,China)

机构地区：[1]太原科技大学电子信息工程学院,太原030024

出　　处：《太原科技大学学报》2024年第4期360-365,共6页Journal of Taiyuan University of Science and Technology

基　　金：山西省重点实验室项目(201805D111001);山西省自动化工程技术研究中心项目(201805 D121020-13)。

摘　　要：针对目前现有方法难以准确识别指针式仪表表盘信息导致读数识别精度不高的问题,研究了一种改进的指针式仪表表盘分割的方法。首先将待分割的指针式仪表图像采用Mask R-CNN网络对仪表图像进行分割,得到仪表中刻度线区域和指针区域的位置,然后选取分割结果中指针区域针尖方向的边缘点作为中心,在原图上构建矩形,采用轮廓跟踪算法对矩形内的图像进行检测,得到图像中指针的区域,最后将两部分指针区域进行融合得到最终的指针分割结果。实验测试结果表明,采用本文方法分割后读数识别结果的误差仅有分度值的四分之一,证明了该方法的有效性和鲁棒性。Aiming at the problem that the reading recognition accuracy is not high due to the difficulty of the existing methods in identifying the dial information of pointer instrument,an improved dial segmentation method of pointer instrument is studied.First will stay pointer instrument image segmentation using Mask R-the network for image segmentation,get the instrument calibration area and the position of the pointer area,then select the segmentation result pointer area pinpoint the direction of the edge points as the center,to build on the original rectangle within the rectangle on the image contour tracking algorithm is adopted to improve the detection,Finally,the two pointer regions are fused to get the final pointer segmentation result.Experimental results show that the error of the reading recognition result is only one quarter of the segmentation value,which proves the effectiveness and robustness of the proposed method.

关 键 词：指针式仪表 表盘分割 Mask R-CNN 轮廓跟踪 指针分割 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]