| In coal mine production environment,gas explosion has always been the biggest threat to the life and safety of underground workers.Gas explosion is mainly a disaster caused by high gas concentration under the mine.Therefore,it is a necessary measure to control the gas concentration and avoid the gas explosion accident to accurately detect the methane concentration.In order to detect methane concentration effectively for a long time,it is necessary to check the methane sensor regularly.This project is supported by the national key R&D project "Metrological technology research of new methane ventilation and dust-proof safety instruments in mines(2017YFF0205500)",the non-dispersive infrared methane transmitters automatic verification system has been studied,which is easy to plug,movable,it can complete the automatic verification of twelve sensors at the same time,what’s more,it can be applied the output signals of various types under complex lighting conditions.The main research contents are as follows.The main research contents are as follows.This topic designs the gas circuit of the verification system,the control circuit based on STM32F407ZGT6,the circuit to detect the current and frequency signals output by twelve methane sensors,and the acquisition,identification and transmission of image signals based on raspberry sensor.To ensure the communication between different function modules in the verification system,RS485 protocol is used for the communication between frequency meter,current meter and single-chip microcomputer,and the communication protocol between single-chip computer,upper computer and raspberry pie is developed.Due to the different power supply voltage and partial isolation of the system,two independent auxiliary power sources are designed to output 15 V and-15 V voltage respectively.The upper computer can complete real-time data display,processing and other functions.In addition,the upper computer can realize human-computer interaction through serial communication,and the verifier can control the whole verification process by manipulating the upper computer.The upper computer mainly displays and stores the verification process data of indication error,repeatability,response time,drift and signal transmission error for the results of printing subsequent verification.Most of the existing calibration devices cannot be moved anywhere and can only be fixed-point to complete the calibration work.Especially recognizing and recording the image signal of the digital display of the sensor,the lower accuracy can not be adapted to work under various complex lighting conditions.Therefore,this paper focuses on the related algorithms of sensor digital display recognition.In the image preprocessing stage,Retinex algorithm is introduced into the field of sensor digital recognition which is mostly used in face recognition and fog removal.It can better find the red part of the digital value when locating the sensor.The multi-scale Retinex(MSRCR)algorithm with color recovery is determined by experimental comparison as the image enhancement algorithm in the preprocessing stage,and good results are obtained.In this paper,character recognition phase is compared to handwriting character recognition.PSO-SVM algorithm based on PCA and convolution neural network algorithm are used to identify the customized dataset of this system.The advantages of the two algorithms are combined,the shortcomings of their respective algorithms are eliminated,SVM is used to replace the classification of convolution neural network to classify and identify the feature information extracted by the convolutional neural network.The combined algorithm is improved because of its long running time.The feature dimension extracted by convolutional neural network is further reduced and then recognized by SVM classifier.The experimental results show that the improved algorithm has higher recognition rate and lower running time.The paper has 67 figures,9 tables,and 72 references. |
PDF Full Text Request