Research On Digital Identification Method Of Substation Digital Instrument

There are a large number of non-intelligent digital meters in the old substation,which cannot automatically synchronize the data detected by the instrument into the substation management system.the substation needs a lot of manpower and material resources to collect the data of the digital instrument and synchronize it to the management system.However,the use of manual meter reading to collect instrument data has many drawbacks,which cannot guarantee the quality of the data,nor can it achieve the real-time detection of the instrument data,which brings a lot of inconvenience to the management of the substation.Therefore,an intelligent method is needed to automatically collect meter values and upload data.The wide application of visual monitoring systems and power inspection robots in substations provides conditions for solving the above problems.With the inspection robot and the monitoring point camera,the digital instrument data in the substation can be collected and uploaded to the management system in the form of images.The paper will study the automatic reading method of digital instrument images collected in substations.Automatically reads from digital instrument images using in image instrumentation,instrument image registration,and instrument representation identification processes.Firstly,the instrument detection method based on YOLO v3 is used to detect the position of the digital instrument in the image and predict the type of the instrument in the area.The algorithm can be calculated in parallel,and the instrument image can be detected in real time under the acceleration of the GPU.Secondly,the image registration method is used to register the detected instrument image to a standard form with the same size and angle,and solve the problem of the difference between the scale and the tilt angle of the image caused by the difference in the shooting distance and angle of the camera.the instrument representation number identification only needs to utilize the image information of the display area.Therefore,the method obtains a projective transformation relationship matrix between the template image and the instrument image to be registered,then maps the pre-marked position of the display area on the template image to the instrument image with registration,and obtains the position information of the display area of the image to be registered,thereby The display area is registered to the standard form.The method can not only reduce the calculation amount of instrument registration,but also facilitate the extraction of characters displayed later.Finally,an instrument representation number recognition method based on prior information is proposed to automatically read the meter display value.The pre-marked position information of the characters that may appear in display area is used to directly extract the displayed single character from the registered meter display area image,the problem that the character cannot be extracted due to the complicated character distribution and other interference image information can be solved.Different segmentation algorithms are used for the characters displayed by different types of meters to reduce the difficulty of character segmentation.The improved Le Net v5 character recognizer is used to identify 0-9 character images and blank character images.Because the accuracy of the numerical data displayed by the five types of digital instruments studied is fixed,this paper bypasses the image processing method and uses the pre-recorded decimal point position to determine the position of the decimal point in the display.