Research On Information Recognition Method Of Substation Instrumentation Based On Machine Vision

With the development of the country and the concept of energy conservation and emission reduction,electric power occupies a dominant position in the energy market.As the core hub of the power system,the substation is an important carrier of people’s production and life.It is very important to ensure its normal operation.The inspection of the instruments in the substation is a key part of it.The traditional manual inspection task is heavy and subjective,which may lead to misreading and missed reading,which is not conducive to the production safety of substations.Pointer meters are widely used in substations because of their low manufacturing cost,strong anti-interference ability,and wide applicability.Therefore,this paper mainly focuses on the pointer meter in the substation,and proposes a method based on machine vision to identify the meter information with a higher degree of automation,higher precision and accuracy.The main research contents of this paper are:(1)The substation instrument detection method based on the Mask-RCNN network model is presented.Aiming at the problem of gradient disappearance and explosion in the deep network in the process of neural network training,an improved Mask-RCNN target detection network model is built.The original backbone network is replaced with Res Ne Xt-50 network,and the recognition and extraction of the instrument panel area is realized by training the instrument image data set.Through comparative experiments,the loss function curve,evaluation target detection accuracy index,recall rate and other data of the improved algorithm and the traditional algorithm are compared to verify the effectiveness of the improved algorithm and improve the detection accuracy.(2)An image enhancement method for collecting instrument images in complex situations is given.Mainly in the case of low light and haze weather,an instrument image enhancement algorithm combining MSR algorithm with color recovery and histogram equalization is given.The fusion algorithm maintains the true color of the image,improves the details of the dark part of the image,and is more suitable for the instrument detection algorithm.(3)The reading identification method of pointer meter in substation is given.In order to save program memory and facilitate subsequent operations,the instrument image is first preprocessed.Aiming at the problem that there is too much information on the instrument panel and it is difficult to identify the pointer and the center of the dial,an improved Canny edge detection algorithm is proposed to improve the accuracy of determining the position of the center of the dial.Secondly,the intersection position information of circle and pointer is detected by Hough transform,and pointer region is obtained by binary image connected domain analysis.After finally refining the pointer and fitting the centerline,select the angle method to calculate the output meter number.(4)Based on the Qt platform,a substation instrument information identification system is developed and tested.The test shows that the system can effectively complete the daily instrument inspection of the substation.