Research And Design Of Electronic Patrol System For Chemical Enterprises Based On Wireless Wide Area Network

Patrol inspections are of great significance to enterprises and communities.The traditional patrol system does not have convenience and real-time performance,and cannot report feedback on patrol information and waste resources in real time.In response to these problems,electronic patrol systems came into being.The popularity of smartphones and the development of NFC(Near Field Communication)technology have brought about tremendous changes in people's lifestyles.The combination of NFC tags and smartphones for electronic patrol systems provides new ideas for patrols.At the same time,in recent years,new technologies such as neural networks and deep learning have developed rapidly,and artificial intelligence has been applied to all aspects of life.As a product of artificial intelligence,CNN has unique advantages in the field of image processing.Therefore,this thesis applies CNN-based image recognition technology to patrol pipeline leak level detection,and develops a low-cost real-time electronic patrol system that supports mobile phone operation.The main work of this paper is as follows:1.Classification of pipeline leakage based on convolutional neural networkIn view of the need for a chemical company in Zhejiang to assess the leakage of liquid pipelines during the patrol process,the technology of automatic classification of pipeline leakage based on patrol photos was studied.Based on the research of automatic extraction of image features by convolutional neural network,this thesis proposes a pipeline leak classification scheme based on convolutional neural network image recognition technology,and designs and implements a pipeline leak classifier based on convolutional neural network image recognition.In view of the fact that the traditional convolutional neural network structure has too many parameters,it is easy to make the network over-fitting problem.A classifier architecture based on GAP global pooling technology is designed to meet the characteristics of pipeline leakage image recognition,which reduces the number of parameters and speeds up.The model training speed reduces the over-fitting problem and improves the classification accuracy.On this basis,several mainstream gradient descent optimization training algorithms for convolutional neural networks are studied.The Adam optimization algorithm is improved,which further speeds up the classifier training and improves the accuracy.In the experimental research,the Keras high-level neural network API of TensorFlow platform is used,and the CNN neural network model is built by python programming,and the CNN-based pipeline leakage classifier is obtained through a large sample set training.The experimental results show that the pipeline leakage automatic identification classifier designed in this paper can replace the artificial pipeline leakage detection better,and its accuracy is over 92%.This thesis describes the process of liquid pipeline leakage image sample acquisition,image preprocessing,convolutional neural network classifier architecture design,classifier training and experimental testing.2.Low-cost real-time electronic patrol system development supporting mobile phone operationIn response to the demand of the patrol system of a chemical company in Zhejiang,the technology of NFC and Android was used to develop an electronic patrol system under the WAN that supports the B/S architecture of mobile phone operation.The system can communicate in real time by means of mobile phone functions,which is convenient for realtime monitoring of management personnel and low networking cost.In the system development,the key information of the database is encrypted to ensure the security of important information;the encryption of NFC tags ensures the security of the patrol information,and can effectively prevent the patrol personnel from forging patrol information;based on the development of Baidu map,It can display the patrol route and the patrol point patrol on the map,making the interface friendly and intuitive.Using the background management system developed by Spring,MySQL,etc.,it can schedule the patrol personnel and manage the patrol data.This thesis describes the whole process of patrol system development,including requirements analysis,overall design,pipeline leakage classification research,functional design,database design,software design and implementation.