Research On Digital Display Instrument Recognition Method Based On Video Capture

With the introduction of China’s smart manufacturing 2025,the trend of intelligent and less humanized manufacturing has become increasingly apparent,and the demand for digitizing instrumentation data in old factories has become stronger.However,instrumentation faces many more realistic problems.First of all,it is limited to technology,process and other factors.There are huge difficulties in the reconstruction of some old instruments.Secondly,the hardware equipment needs to be shut down for a long time during the transformation of the instrument,which not only requires a large amount of investment,but also loses huge economic benefits.Therefore,the method of instrument identification based on video capture can ensure that the old instruments are not digitized and that they do not stop work during the transformation,thereby avoiding economic losses during the transformation,which has practical significance for practical engineering applications.Based on the actual engineering application scenario,this paper proposes and implements a universal instrument identification framework.The instrument identification process is unified into three sub-processes: instrument area detection,instrument feature segmentation and extraction,and instrument reading identification.According to different instrument types,different algorithms are applied to each sub-process,which can have a significant effect on the digital display instrument,disc pointer instrument,liquid column instrument and other types of instrument reading recognition.In the recognition of digital display instruments,Faster R-CNN is used to realize the automatic detection of the instrument area,and then a projection method based on the combination of horizontal projection and vertical projection is proposed and realized.It can eliminate various noises in engineering applications for characters.The impact of segmentation and recognition.Finally,based on the convolutional neural network,a Led Net was used for digital character recognition,and the readings of the digital display instrument were finally obtained.In the recognition of the disc instrument,Faster R-CNN was also used to implement the instrument area.Automatic detection,in the stage of instrument feature segmentation and extraction,Huff detection was applied to realize the feature extraction of the dial and hands,and finally the relative position of the dial and pointer was used to calculate the reading of the disc meter;finally,in the identification of the liquid column instrument In the above,Faster R-CNN is also used to realize the instrument area detection,and then the difference comparison,relative position and range information of the liquid column area are used to calculate the liquid column instrument reading.In addition,this paper designs and implements a set of image data acquisition methods for engineering applications.In summary,this paper proposes a universal recognition framework,and based on this framework to realize the reading recognition of industrial display instruments,there is great engineering application value in actual engineering applications,and there are certain optimizations in the numerical character segmentation algorithm And application innovation.In the extended application of the above framework,the reading recognition of disc meters and liquid column meters is realized.In order to solve the problem of image data acquisition in engineering applications,a set of engineering data acquisition methods are designed and implemented.