Pointer Type Instrument Intelligent Recognition System Based On Inspection Robot

In the process of industrial production,meters are effective devices for monitoring the status of equipment.Pointer instrument for its strong anti-interference ability,easy maintenance and other features,has a wide range of applications in the industrial field.The traditional method of manually reading the number is time-consuming and laborious.With the rapid development of inspection robots,inspection robots based on computer vision technology have gradually been applied to the work of equipment and instrument inspection.Aiming at the industrial environment with low level of intelligence,more equipment to be inspected and wide distribution,this paper designs an intelligent recognition system for pointer meters based on inspection robots.The system has the functions of patrol target detection,pointer type instrument representation number recognition and motion blur image restoration,is able to realize intelligent detection and reading recognition of equipment and meters.Inspection target detection: Use Yolov5 deep learning algorithm to detect instrument targets in the field of view.In order to avoid the failure of recognition due to the small target of the instrument,firstly,the instrument box is inspected for rough positioning of the instrument,and then the instrument target is detected by the camera focusing.After successful detection,feedback on the position of the instrument,adjust the camera pan/tilt so that the target is located in the center of the field of view to improve the recognition accuracy.Pointer type instrument representation number recognition: According to the detection efficiency and accuracy requirements,the traditional image processing method is used for the instrument representation number recognition.After the camera calibration is completed,according to the specific characteristics of the field instrument,the corresponding image preprocessing method is selected for dial extraction and pointer detection,and the reading result is obtained by judging the position of the pointer in the display interval.Motion blur image restoration: In order to realize the recognition of the number of the instrument in the moving process,and improve the inspection efficiency.Using the motion blur degradation model,the image is subjected to noise analysis and blur kernel estimation,and the restoration algorithm is used to process the image motion blur according to the obtained results.The development of the visual inspection system in this paper is based on the Ubuntu 18.04 operating system,using the ROS platform and the Pytorch deep learning framework.The system uses the video collected by the inspection robot to make an image test set and perform performance tests.The results show that the system can meet the effect of the accuracy of the instrument display number recognition is higher than 95%,the target detection time is less than250ms/frame,and the instrument display number recognition time is less than 150 ms/frame.It can efficiently and accurately complete the task of meter detection and display recognition.