Research On Digital Meter Reading Method Based On WSN And CNN

Wireless sensor networks(WSN)are widely used in many fields,such as smart homes,environmental monitoring,urban traffic,and machinery monitoring with the advantages of low cost,low power consumption,high reliability,convenient layout,and maintenance.Convolutional Neural Network(CNN)could achieve an excellent performance in image processing,natural language processing,and other fields due to its structural features,such as local area connection,weight sharing,and downsampling.Digital meters are widely used in industrial production and daily life.However manual meter reading is time-consuming and laborious.It is of great significance to look into the automatic meter reading technology of digital meters combined with wireless sensor networks and convolutional neural networks.This paper combines wireless sensor network technology with convolutional neural network,which proposes a digital meter reading method based on WSN and CNN.The household digital water meter is selected as the research object.And a digital meter reading system is designed on the basis of WSN.On the wireless sensor terminal node,functions such as the collection,processing,and value recognition of the digital instrument image are realized.Through the wireless sensor network,only the recognized value of the meter is transmitted to the computer,which achieves the remote automatic meter reading of the digital meter in operation on the spot.The specific work of this paper is as follows:1.This paper summarizes the research status and development trends of wireless sensor networks at home and abroad.The related recognition algorithms of digital meter readings is analyzed.To embedding wireless sensor network nodes,the suitable image processing algorithms are selected.The paper also builds lightweight convolutional neural networks.Here MATLAB platform is applied to conduct the simulation of CNN.2.The hardware platform construction of the digital meter reading system based on NXP's JN5169 development board is introduced in this paper.The appropriate image acquisition module and LCD are selected for the capture and display of meter images.3.On the basis of the Beyond Studio for NXP integrated development environment,this work conducts the image preprocessing,image filtering,image segmentation and recognition algorithm program development of the digital meter reading system as well as the wireless network debugging.4.Under laboratory conditions,WSN and CNN are performed in a digital meter reading experiment.The experimental results demonstrate that the designed wireless sensor network terminal node is able to collect digital instrument images.And the CNN algorithm is performed to complete the digital instrument value recognition process on the terminal node.Then,the indicated value could be transmitted to the wireless sensor network through the wireless sensor network.Finally,the indicated value will be transmitted to the computer via serial communication via the coordinator node.