Reading Of Pointer Instrument In Substation Environment With Machine Vision

In this paper,the pointer instrument of converting station with complex background is taken as the research object,and.based on image processing,the reading recognition algorithm is designed under machine vision.Aiming at the particularity and typicality of pointer instrument,an effective target extraction,scale reconstruction and indicator reading algorithm are designed.The particularity of the pointer instrument itself is mainly embodied in the circular,square and fan-shaped surface of the dial,which has different shapes.Its typicality is mainly embodied in the composition structure of the pointer instrument,such as the single-head rotating pointer,the scale line with circular interval distribution,and the digital mark closely adjacent to the scale line.However,due to its complex installation environment and long use time,as well as the limitation of shooting conditions,the image quality of the instrument is poor,which seriously affects the readability of the indicator,and reduces the efficiency of automatic inspection on the power grid side.In view of the above characteristics,this paper designs an effective processing algorithm for each link in the process of identification of five types of dials:pressure gauge,thermometer,ammeter,voltmeter and power meter.In order to solve the problems of complex installation environment,unsatisfactory shooting conditions,unclear dial and low contrast,a multi-direction edge monitoring algorithm based on Canny operator is designed in the pre-processing stage.The algorithm can effectively suppress irrelevant details,highlight the main contours to extract the dial accurately,and remove background equipment,sky and other areas.I the phase of disk extraction,the center of rotation is determined by disk filling and huogh circle detection,and the pure disk diagram and the center of disk architecture are obtained.Aiming at the situation that the disc calibration is not clear or defective,a method of calibration location and calibration reconstruction based on Fourier transform is designed by making full use of the typicality of the calibration.This method transforms spatial analysis into frequency analysis,which mainly includes the following aspects:1)Construct a multi-factor interpolation mapping model.The model maps the annular distribution of scale to linear distribution,improves the scale aggregation,and facilitates the description of scale.2)Design the frequency domain analysis method to locate the calibration band.The mapping image is segregated orderly and Fourier transform is performed to define ordered region energy.The calibration band location algorithm is designed by analyzing the corresponding relationship between the mapped disk and the regional energy curve.This localization algorithm integrates scale,number and pointer,which greatly reduces the computation of step-by-step feature extraction.When the maximum allowable error is 0.05,that is,the overlap rate is greater than 90.91%,the positioning success rate of this method is more than 90%.3)Design scale reconstruction model.The model describes the one-to-one correspondence between scale and numerical value,which can effectively reduce the influence of scale missing and scale uneven distribution on reading.Finally,the number of dials is identified by reconstructing the model and the relative position of the pointer in the mapping image.The design of this subject has achieved good recognition effect in practical application,and has strong universality.It can be used to read the indicators of dials with different shapes.