Research And Implementation Of Automatic Reading System For Pointer-type Meter

Pointer-type meter has been widely used in people’s production and life.At present,the reading of pointer-type meter mainly depends on human force,which is seriously restricted by human strength and measuring environment.With the development of image processing and computer vision technology,the pointer-type meter automatic reading technology based on image processing method has become a hot research topic.This paper is from the national major instrument project "near space high speed target plasma electromagnetic science experimental research device".Aiming at the problem of automatic reading of power supply instrument of inductively coupled plasma device in this project,this paper focuses on the automatic identification and positioning of meter area,geometric correction of meter image,pointer detection and reading calculation.On the hardware platform of the monitoring system in the laboratory where the device is located,the software system of automatic reading of pointer ammeter and voltmeter is designed and implemented,and it is successfully applied to the multi meter automatic reading environment in the laboratory.The main contents of this paper are as follows:1.In order to identify and locate the area of pointer-type meter in different illumination conditions and low resolution images,convolutional neural network is used to identify and locate the instrument automatically.Firstly,the collected instrument image is labeled as training data set and test data set.Then,Yolov3 is used to train the training data set to obtain the weight coefficient of the network model.Among them,the parameters of the prediction box are modified to make the prediction box completely cover the instrument area.Finally,after testing,the method can automatically and correctly output the coordinates of the rectangular area of the instrument under different illumination conditions.2.The geometric correction of the image by using the homography matrix is proposed,which solves the geometric distortion of the rotation angle of the pointer in the image,and can restore the true rotation angle of the pointer.Firstly,the edge of square instrument frame is refined based on skeleton extraction algorithm.Then,the four lines of the instrument frame are extracted by multi-linear clustering algorithm.The vertex of square instrument is obtained according to the intersection of the lines,and corresponding to the real vertex coordinates,the homography matrix from the image plane to the world plane is calculated.Finally,the homography matrix is used to correct the pointer line and obtain the true angle of the pointer rotation.3.In order to solve the special requirements of the laboratory for reading speed,this paper does not adopt the dial scale recognition algorithm with high time complexity and fine pointer extraction algorithm,but adopts frame difference method to extract the rotation position of the pointer,obtains the rotating pointer image by positive and negative frame difference data,and then calculates the straight line of the pointer by single line fitting algorithm,In order to reduce the noise,RANSAC method is used to improve the fitting accuracy.4.Based on the above principles and algorithms,this paper implements a pointer type instrument automatic reading software system based on the monitoring camera.Among them,the SDK function provided by the manufacturer is used to access and control the camera;The video of three surveillance cameras is read and processed in real time by multithreading;The recognition results of neural network are transmitted to the automatic reading software by file;And the Open CV function library is used to realize the relevant image algorithm.The actual test shows that the speed of the software system can reach 10 frames / s when three cameras detect 12 pointer instruments at the same time.Under the condition of low image resolution of the monitor camera,the error of the software system is less than 5%.The software solves the problem of automatic reading of power supply instrument in laboratory,and meets the requirements of the system.