Research On Automatic Reading Recognition Of Mechanical Instruments

Due to its simple design structure and cheap cost,mechanical analog measurng instruments are applied to a variety of fields,such as mine,oil extraction,Power station etc.However,compared to the traditional way of obtaining analog measurng instruments recognition information by artificial reading,the application of Computer Vision Technology(CVT)has greatly reduced the labor cost and improved the automation detection level of analog measurng instruments,which shows great development potential and broad application prospect.In this paper,mainly based on the theory of digital image processing technology,we put forward an automatic reading recognition method for mechanical analog measuring instruments based on the combination of center projection and Hough Transform of practical significance.The main contents of this paper include three aspects: first of all the image preprocessing has been discussed,including image graying,binarization,smoothing and sharpening and so on.After that the principle of image line detection has been researched,which including the pointer line detection principle of center projection and Hough transform.Besides the work such as calibration of center of rotation,needle orientation and confirmation of effective range has been accomplished as well.Last but not the least,the principle of automatic identification for mechanical analog measuring instruments readings has been studied,the dail reading results have been calculated according to the linear proportional relationship between the pointer and the measurng instruments' range.Later error analysis of automatic reading recognition has been classified and discussed based on the experimental results.In ord to ensure that that effective range circular arc of the dial is more closer to the circle,Image Correction Technology(ICT)has been adopted to redress the geometric circle structure of instrument image in the process of image preprocessing.And then adaptive median filtering method has been used to gray image filtering.Contour tracking method has been not only used to segment dial area and pointer area but also denoise the whole image.In the process of line detection,Hough Transform is adopted to calibrating the center of rotation and the effective range of dial,and then center projection is used for fast line detection.So that the pointer line detection method of this paper is principally based on the combination of center projection and Hough Transform,and compared the advantages and disadvantages of only use Hough Transform method and the combination of center projection and Hough Transform method.In the analysis of the experimental results,the error and solution of the multi-pointer instruments are analyzed in detail,for example the pointers are coincident or parallel or points to the scale line etc.The experimental results showed that the error of center projection combined with Hough Transform method is within 5 %.Considering the reliability of the results,the time cost of this method mentioned above is only about 7.0 ms for single pointer and 16 ms for double pointers.On the whole,the detection speed of this method is less than 0.1 times that of Hough transform method,which has reliable application value and development prospect.