Research On Digital Meter Reading Recognition System Based On Wireless Sensor Networks

At present,wireless sensor networks technology belongs to the high-speed development stage,which is widely used in crop monitoring,fire safety detection,intelligent life,intrusion monitoring and other fields.Wireless sensor networks can be used not only to measure scalar value or dynamic signal,but also to identify complex information such as image.As a common digital meter in industrial production and daily life,the reading mode of digital meter is becoming more and more intelligent.For some places,the replacement of intelligent meter is not only a large amount of construction and high cost,but also the camera direct reading automatic meter reading technology can solve this problem well.Based on wireless sensor networks technology and image processing technology,this paper designs and builds a digital meter reading recognition system with wireless node as the core,which realizes image acquisition,image processing and character recognition on the terminal node.According to the characteristics of the research object,specific problems are analyzed to meet the requirements of the system Practical application requirements.The specific work of this paper is as follows:1.Summarize the research progress and application of wireless sensor networks in image recognition in recent years,and analyze the constraints of wireless sensor networks nodes in the face of image information.Complete the design of digital meter reading recognition system,select JN5168 wireless node as the core hardware of the system recognition terminal,select ov7670 image sensor as the acquisition module,and complete the hardware system construction.2.Analyze and compare the different algorithms through the simulation experiment,and determine the image processing algorithm and implementation flow of the meter reading recognition research.Complete the system software design,image acquisition,image processing and other related program development,complete the experimental results analysis process to display the processing results of the serial communication protocol.The training process of BP neural network is completed by MATLAB,and the character recognition of BP neural network is realized on the end node.3.Complete the test work of the wireless sensor networks reading recognition system,use a large number of meter images collected to test the recognition rate of BP neural network,analyze the factors that affect the recognition rate of the system,and test the effect of different environmental factors on the image processing effect and recognition rate of nodes through experiments,which verifies the applicability of the reading recognition system designed in this paper.The experimental results show that the recognition rate of the whole character and the rolling character is 97.4% and 84.5% respectively in the normal indoor environment.In addition,compared with the traditional way of collecting and transmitting the original image to the upper computer,this method only needs to transmit the final recognition results,which solves the problem of the limited communication bandwidth of the nodes.