Research On Recognition Algorithm Of Instrument Panel Display State

The automatic identification system of meter indication is an automatic system that combines optics and electromechanics.It can collect the image of the instrument through the camera without changing the original system,and obtain the indicator information of the instrument by locating and extracting the target.The automatic identification system of the instrument can improve the informatization and intelligence of the detection device and the system,and it also involves machine vision,image engineering,artificial intelligence,micro-electromechanical system and other interdisciplinary fields.The existing automatic instrument identification system is mainly oriented to substations,and the instrument images of the substation are collected by automatic inspection robots.Therefore,the research of automatic instrument reading system is of great significance.This paper is based on the YOLO deep learning network to study the automatic meter identification algorithm.First,establish a hardware system and enhance the image.Build the hardware system,analyze the environment of the substation,simulate the environment of the substation,add compound noise including Gaussian noise and salt and pepper noise to the collected images,and design a multiple weighted approximation filter algorithm based on this noise.The received image is processed so that the processed image can suppress the noise while retaining more details in the image.Secondly,extract and segment the instrument features.The YOLO network is used to detect the image of the instrument panel,the instrument panel data set is established and the YOLO network is trained through the training set,so that it can locate the instrument in different natural environments,and at the same time,the positioned instrument is segmented and appears The distorted instrument image is corrected by perspective transformation.Then,the recognizer indicates the number.Analyze the classic method of identification of each instrument,adopt the line scanning method to determine the indication for indicator instruments;for digital display instruments,adopt the threading method improved by adding constraints;design a deep volume for pointer instruments The product network extracts the pointers and calculates the indication of the meter through the extracted pointers.Finally,through experiments to verify the feasibility of the indicator recognition algorithm structure of the instrument panel designed in this paper,and design software to encapsulate the steps appearing in this experiment so that it can achieve the result of normal instrument identification.