Design And Implementation Of Automatic Sorting System For Dismantled Electric Meters Based On Machine Vision

Electric meters are electrical equipment in people’s life.However,the number of dismantled electric meters every year due to reasons such as rotation and dismantling is huge.In Sichuan Province alone,the number of dismantled electric meters in one year amounts to one million and six hundred thousand.At present,the sorting of dismantled electric meters in the industry mainly relies on manual sorting,which has low work efficiency and high labor costs.Therefore,it is necessary to study the realization of automatic sorting of dismantled electric meters,instead of manual sorting.At present,the relevant research is mainly aimed at the barcode scanning and identification of the dismantled electric meters.The parameter information of the meter model,version,manufacturer,and electrical specifications has not been identified and extracted,and a large amount of valid information is missed,which is not conducive to the subsequent analysis of relevant data of dismantled electric meters.Such as quality analysis,problem location and traceability,etc.In response to the above problems,this article has developed a machine visionbased automatic sorting system for dismantled electric meters,which can be split into two sub-modules,a mechanical platform and an image recognition software.(1)Mechanical platform: It is composed of image acquisition device and sorting device.The image acquisition device fixes the dismantled electric meters on the assembly line and then shoots,acquires the meter image,inputs the image into the software system,completes the extraction of the corresponding parameter information,and transmits the information to the subsequent sorting device;the sorting device is based on the extracted information transport the meter to different assembly lines to complete the automatic sorting process;(2)Image recognition software: The main task of the software is to identify the character information of the meter image and complete the extraction of the corresponding parameters.Based on template matching,the coarse positioning of the character area is realized,and the image of the electric meter is divided into multiple area ROIs,which are input into different subsequent modules.Based on the gray features of the ROI image and the Ada Boost classifier,the meter version is classified for the lower area ROI,and the open-source barcode recognition tool is used to achieve barcode recognition for the middle area ROI.Based on CTPN,the precise positioning of the characters on the middle area ROI is realized,and the precise positioning coordinates of each line of characters are obtained,and the coordinates are input to the character recognition module.The method of synthetic simulation data set is used to train the recognition model based on Dense Net+CTC to complete character recognition.Use regular expressions to extract the required meter model,version,manufacturer,electrical specifications and other parameter information from the recognized character information.Finally,use the Tkinter tool to package the software to complete the design and implementation of the humancomputer interaction interface.In order to test the feasibility of the automatic sorting system,the experiment selected 1,000 meters to test the system.The test results showed that the average recognition accuracy of the system reached 98.4%,and the detection time of a single meter was 1.56 seconds,which met the project requirements.This paper develops a machine vision-based automatic sorting system for dismantled electric meters.The system uses image acquisition devices to obtain meter images,and then uses OCR models to complete the character location and recognition in the meter images,and complete the meter model,version,and manufacturer,electrical specifications and other related information extraction.Finally,based on the identification information,the dismantled electric meters is transported to different assembly lines to complete the sorting of different categories.The system has complete functions and strong robustness.The information content that can be extracted exceeds the existing identification system of the dismantled electric meters,which is conducive to the subsequent data analysis of the dismantled electric meters,such as the quality analysis of the company’s production meters,and the location and traceability of abnormal electricity usage problems,etc.It can replace the manual sorting of dismantled electric meters,realize the automatic sorting process,improve work efficiency,and improve the sorting efficiency of dismantled electric meters industry.