Machine Vision Based Design And Implementation Of Embedded System For Meter Recognition

Many production equipment running on site in manufacturing industry are mostly mixed with old instruments.The acquisition and dumping of new instrument operation data is more convenient and quick.For some old instruments for many years,there is no data access interface and is not convenient for unified data management.For some abnormal alarms,it depends on manual reflection;for some production data,it depends on manual recording.If the instrument is upgraded and replaced,the whole body will be affected,and it will face greater expenditure costs and even various production and operation risks.Based on the practical application project of Master of Engineering,this paper designs and implements the embedded system of instrument status recognition based on machine vision.It uses image processing and analysis technology combined with industrial IoT technology to take pictures of the current machine display panel through a camera,or display screen or pointer instrument or indicator light at manual operation panel are used to monitor the operation data and running status of the machine in real time.After analysis and processing,the machine operation data and running status record are saved in the database,and linkage control is realized with other related systems.Users can also access the mobile phone and computer web page to get the running status of the machine in real time.The main work is as follows:Firstly,the quality of light imaging is solved by lighting technology from the perspective of imaging.The distortion correction algorithm is used to correct the distorted images taken from various perspectives,and then the image noise reduction algorithm is used to filter out the particle noise and enhance the detailed characteristics of the image.Secondly,the recognition algorithm is described for four typical application scenarios,which are state indicator,pointer meter display,digital tube display and text character.For the identification of the status indicator and the number of the digital tube,the necessary identification parameters can be set by the user’s upper computer,which can significantly improve the versatility of the system.For the recognition of pointer instrument indication,a line angle voting algorithm based on DDA line drawing algorithm is proposed.It can ensure the processing speed and increase the stability of recognition.For the mainstream text character recognition CRNN+CTC algorithm,the application in the current scene is improved.The convolution part of feature extraction is replaced by lightweight MobileNetv2,so as to improve the detection and calculation speed of edge device.Thirdly,the embedded system of instrument status recognition is introduced from the point of view of software and hardware design.Among them,the upper computer parameter configuration software specifically introduces the software development environment and development platform,and combines the software screen to introduce the design of the human-computer interaction function.About the embedded system,it introduces the selection basis of the main control module and image acquisition equipment,and designs and implements its software program.The instrument status recognition method proposed in this paper combines embedded equipment and host computer software to form a general terminal detection system,which realizes the accurate detection,image correction and recognition of various instruments in general scene.Compared with the existing traditional image processing algorithm and deep learning algorithm for the identification of single type or some specific type of instrument in a single scene,The reading recognition accuracy is higher,the robustness is stronger,the universality is better,and the industrial application prospect is good.