Research On Recognition System Of Pointer Meter Reading Based On Machine Vision

The pointer instrument has been widely used in the process of industrialproduction, because of its advantages such as its simple structure, high reliability, lowprice, high legibility. The use of machine vision technology has an important practicalsignificance and research value to realize the automation of the dial instrument’srecognition process and to improve the accuracy and the efficiency.First of all, this paper selects hardware equipment for the image acquisitionsystem by analyzing the accuracy requirement of the instrument, the imageacquisition process of the instrument and the relevant equipment. Then, basing on theanalysis of the instrument’s feature which has been already collected and theillumination mode, this paper develops a side face beaming scheme with bilateralcolumn light source, and gets an initial target image with higher quality. Next, tohighlight the instrument’s information, this paper preprocesses the initial dialinstrument’s image by using the digital image processing technology. Then, this paperapplies Otsu algorithm to select the optimal threshold for processing the imagebinaryzation; and employs the relevant morphological algorithms to process andstrengthen the feature information of the instrument image.Second, in the traditional algorithms of the dial extraction, it is difficult tochoose Hough algorithm’s parameters; the polar coordinate parameter cannot reflectthe related information of line parameter; the least square method does not have wellglobal quality and it is easily affected by noises. So this paper puts forward animproved least square algorithm using RANSAC to fit the pointer line. Thisalgorithm is robust and its fault-tolerant ability is very strong, and it can obtain apointer line with higher credibility. Considering the identified instrument’s ownspatial position, this paper extracts the scale gravity center to dial center fitting byusing the gravity method, which can solve the problem that the pointer center in notgenerally coincide with the dial center in the actual situation. Finally calculate pointerdeflection to realize the pointer meter reading and recognizing Affected by the perspective projection distortion in the application of visualdetection technology to identify the spatial position of the instrument, the actualimaging model of the instrument is a nonlinear model. This paper analyses theimaging model of the camera; demarcates the instrument’s plane based on the cameracalibration principle, establishes a local world coordinate system on the instrumentplane, and reconstructs the instrument.Finally, this paper uses an experiment to test the accuracy of the reading andrecognizing method rose in this paper. The experiment results shows that this methodcan satisfy the requirements and needs of the actual measurement and application.