Design And Implementation Of Intelligent Digital Display System For Pointer Instruments

With the rise of Industry 4.0,enterprises have an increasingly intense demand for digitalization of instruments in the production process.However,a considerable part of the existing large and expensive pointer instruments in these industrial enterprises are still in the process of transformation,and they do not have the functions of automatic data reading and remote transmission.Huge renovation costs.However,relying on manual readings and transcriptions has problems such as discontinuity,inaccuracy,and inaccessibility of workers.In order to promote the transformation of traditional industry 4.0,continue the use value of these old pointer meters,and facilitate the digital management of enterprises,this paper studies the theoretical method of intelligent digital display for pointer meters based on the deep learning framework,and designs and implements an intelligent digital display system.The main research work is as follows:1.The establishment of field instrument data set and pointer scale data set.In view of the problem that there is no standard pointer instrument data set at present,first of all,the pointer instrument images in a natural gas purification plant,an oil base and the network are collected and organized,and 1000 on-site instrument pictures are produced.According to the processing requirements of the standard data set,Rename it;then use Labelimg and Labelme two labeling tools to label the instrument panel area and pointer scale area involved in each photo one by one;finally,based on the Python language platform,the produced instrument,pointer scale The two data sets are processed into the required.txt and.png file formats respectively,and the instrument data set and scale pointer data set of this paper are established.2.Research on the positioning method of pointer instrument in complex industrial background based on YOLOv5s model lightweight method.The background of the on-site instrument image is complex,and the recognition accuracy is affected by factors such as illumination and angle.Based on the YOLOv5s framework,this paper builds the model structure of the pointer-type instrument panel positioning.After the parameters are optimized,the trained model is obtained.Due to the poor real-time and portability of the network,this paper further completes the lightweight design of the YOLOv5s model,and performs channel pruning on the BN layers of the Back Bone and Neck parts of the network.Model as a pointer gauge positioning method.Compared with YOLOv5s,the positioning time is reduced by 18.18%and the model size is reduced by 56.35%under the condition that the positioning accuracy remains unchanged,which provides a basis for reducing the running time of the entire system.3.U~2-Net-based pointer meter reading recognition method and its realization.First,the image is preprocessed to solve the interference such as image tilt and dark light.In view of the low detection accuracy of the pointer scale,which is easily affected by occlusion and blur,this paper uses the U~2-Net model to segment the pointer and the scale,and the segmented two-dimensional image is divided into two-dimensional images.Convert it into a one-dimensional array,locate the relative position of the pointer on the scale,and use the distance method to calculate the reading to obtain the meter reading result.After experimental testing,the test accuracy rate is 93.42%,the average absolute error rate is 0.0262,the average reading error rate is 0.597%,and the average test time per image is 0.402s,which meets the actual needs of the industry.4.Design of intelligent digital display system for pointer instrument.Combining the above YOLOv5s lightweight positioning algorithm,meter reading algorithm and Py Qt5,and monitoring the camera through an external USB camera,the intelligent digital display human-computer interaction system of the pointer meter is designed and implemented.The system mainly includes two parts:the login part and the main interface.The login interface of the system uses the database to verify whether the login information is correct to ensure the security of the system;the design of the main interface mainly includes preview tab pages,display tab pages and visual tab pages.Through an external USB camera,the collected images are used as input,and the model trained before the system is used to perform automatic instrument positioning,which can realize the current acquisition of instrument images,local images,and custom acquisition image time intervals.Intelligent digital display and visualization of historical data results.