Study Of Automatic Value Identification Of Pointer-type Pressure Gauge Based On Machine Vision

This paper presents a new method of value identiifcation of pointer-typepressure gauges based on machine vision, by which an automatic calibration systemcan be designed. Pointer-type pressure gauges can’t be replaced because of the nicestability and the high accuracy. They are widely used in the industrial production andthe detection process. Because of the limitation of human eyes，a man can onlyidentiyf the pointer position roughly when it locates between two scale lines.Different results will be obtained to the same value by two operators. Unreliableresults will be occurred because the repetitive action of reading panels is a tiringwork. In the process of veriifcation, the error caused by human factors would begreatly reduced by using machine vision and image processing technology. Thismethod has a strong objectivity compared with the human method.The value of pointer-type pressure gauge is obtained with the help of machinevision and image processing technology. A image acquisition system is build to get ahigh quality image by a industrial camera which its resolution is1280X1024. Thispaper use image subtraction, binarization, image thinning to make the feature of theimage more obvious. Hough transformation is used to detect the lines in theprocessing of extracting the radius and the rotating center of pointer and detect thepointer in single image. Combined with the angle and distance method, a newmethod is proposed to compute the value of the gauge. During the process ofrecognition, the algorithm of removing the pixels data outside of the circumference ofthe Pointer before Hough transformation when detecting the angle of pointer in asingle image greatly reduces the impact of the unrelated pixels so that the inlfuenceof redundant pixels which are useless for image recognition is reduced.The job in this paper is completed and run at the platform of Windows.Considering the compatibility issues, Visual C++6.0is applied to program theimage processing algorithm. Experiments are carried out on a pressure calibration device. The appropriateimage processing eliminates the redundant pixels to decrease the influence of thenoise pixels. The experimental results show that this machine vision method ofreading the pressure values on the gauge panel automatically is reliable.