Research On Reading Method Of Pointer Meter Based On WSN And CNN

Wireless sensor networks(WSN)have many advantages such as low cost,no cables,easy deployment,convenient maintenance,etc.,and are widely applied in environmental monitoring,industrial equipment monitoring,medical and health,smart home and other fields.The pointer meter is widely employed in the modern industrial process due to its inherent advantages such as low cost,high reliability,and simple structure.The automatic interpretation method of the pointer meter in the existing literature basically needs the support of a wired system or needs to run on a robot with abundant hardware resources,which makes the installation and operation cost higher.This paper takes the pointer pressure gauge as the research object,studies the automatic reading method of pointer meters based on WSN and CNN,combines image processing technology and convolutional neural network algorithm to design and realize the automatic reading system of pointer meters.After the end node collects the pointer meter image,it first processes the meter image in the order of grayscale,edge detection,edge refinement,dial positioning and pointer extraction.Then input the pointer coordinates into the CNN model,and use the CNN model to classify to get the large scale value.Finally,calculate the angle value of the pointer deviation from the large scale.The sum of the large scale value and the deviation value is the value indicated by the instrument.After the end node completes the meter reading,the indicated value is sent to the coordinator node by wireless communication,and the coordinator node receives the data and sends it to the upper computer for display through the serial port.In this paper,a structure-optimized convolutional network model is used to enable it to run on n odes with limited resources.The specific work content is as follows:1.Summarize the pointer meters interpretation methods in the existing literature,and use MATLAB simulation tools to simulate and test image processing related algorithms and convolutional network algorithms.2.Build the JN5169 basic hardware platform and its software development environment,the main hardware includes end nodes and coordinator nodes,design and debug the basic image acquisition program of the end nodes,use the end nod es to collect a large number of pointer meter images and generate training samples.3.Develop an automatic reading software system for pointer meters based on WSN and CNN,including basic programs for related modules,pointer meter reading programs,wireless network transceiver programs for end nodes and coordinator nodes,and host computer software design.4.A series of experimental tests were carried out on the proposed algorithm in a laboratory environment.The experimental results show that the pointe r meter reading algorithm in this paper is feasible.The system has a certain degree of robustness to interference,and can identify different meters,and the identification reference error is less than 0.3%.The end node only transmits the indicated value,which alleviates the problem of limited communication bandwidth of the node.