基于特征间几何约束的指针仪表读数识别方法

Pointer Instrument Value Identification Method Based on Geometric Constraints among Features

在指针式仪表示值监控系统中,倾斜拍摄仪表图像时会导致识别结果出现偏差,为此提出一种基于仪表特征间几何约束的识别方法。首先利用一幅平铺于仪表上表面的平面标定板图像实现相机参数标定,生成仪表在标定平面上的虚拟图像;然后利用表壳上表面、刻度平面、指针平面与虚拟像平面间的平行关系,确定表壳圆的圆心、仪表刻度圆心、指针转轴中心特征间的透视投影约束关系,利用图像的极坐标变换和离散傅里叶变换提取仪表刻度圆心的最优位置及起始刻度线的角度,利用Hough变换的投票原理估计指针的转动角度;最后利用角度法实现仪表读数的解算。以某型号指针式温度计为测试对象进行实验验证,识别结果的平均误差为0.23℃,最大误差为0.7℃,本文方法能够准确识别仪表读数且识别精度优于人工读数。

In pointer instrument value monitoring system,in the case that the instrument image is sideways captured,the scale pointed in the image might deviate from the true.A pointer instrument value identification method based on geometric constraints among features was presented to address this problem.Firstly the camera parameters were calibrated from an image of a planar calibration pattern which was put on the top surface of the instrument.The instrument’s virtual image on the calibration plane was created.From the parallel property among the instrument shell’s top plane,the scale plane,the pointer plane and the virtual image plane,the perspective projection constraints among the center of the shell circle,the center of the scale lines,and the rotation center of the pointer can be obtained.Next the optimal locations of the scale center,the angle of the start scale line were extracted from polar coordinate transform and discrete Fourier transform of the image.The pointer rotation angle was estimated from the voting principle of Hough transform.Finally the instrument value was computed by the angle method.The method was tested on a pointer thermometer.The average error is 0.23℃,the maximum error is 0.7℃.The experimental results showed the proposed method can identify the value correctly,and the accuracy is higher than manual reading method.