摘要
传统的二次仪表校准方式多为人工校准,校准结果不但误差大,还工作效率低、人力成本高。主要针对温度二次仪表中常用的指针式仪表示值读取进行自动校准研究,采用YOLOv3的目标探测算法对复杂背景下的仪表进行自动校准,利用MobileNetv3的特征点检测算法对仪表上的刚性特征点进行校准,并通过透视变换得到仪表的正视图。针对仪表正视图中存在的干扰问题,通过一系列的预处理方法解决,如对比度增强、图像降噪、形态学处理。试验结果显示,该方法在测验集方面与人工校准的相对误差小于0.1,引用错误小于5%,且具有较高的辨识准确力,适用于温度二次仪表示值读取的自动校准。
The traditional secondary instrument calibration method is mostly manual calibration,and the calibration results are not only error,low efficiency and high labor cost.It mainly focuses on the automatic calibration of the pointer meter readings commonly used in temperature secondary instruments.It uses YOLOv3's target detection algorithm to automatically calibrate the instrument under complex background,uses MobileNetv3's feature point detection algorithm to calibrate the rigid feature points on the instrument,and uses perspective transformation to obtain the front view of the instrument.Aiming at the interference problem in the front view of the instrument,a series of preprocessing methods are used to deal with it,such as contrast enhancement,image noise reduction and morphological processing.The test results show that the relative error between this method and manual calibration is less than 0.1,and the reference error is less than 5%,and it has a high identification accuracy,which is suitable for automatic calibration of temperature secondary instrument readings.
作者
范莉
Fan Li(Jiuquan Metrological Testing and Verification Institute)
出处
《上海计量测试》
2023年第1期43-46,53,共5页
Shanghai Measurement and Testing
关键词
图像处理
温度二次仪表
示值读取校准
透视变换
image processing
temperature secondary instrument
reading calibration
perspective transformation
