Research On Gas Meter Reading Recognition Based On LeNet-5 Model

With the continuous advancement of urbanization in China,the demand for gas among urban residents is increasing,which undoubtedly increases the workload of meter reading and billing for staff.The meter reading methods of traditional gas meters and IC card gas meters are no longer suitable for today’s intelligent life.Therefore,domestic and foreign scholars have developed remote meter reading systems,but most of them have changed the structure of gas meters.Moreover,this meter reading technology lacks reliability,flexibility,and universality,making it difficult to promote it on a large scale.To quickly and accurately identify gas meter readings and achieve self-service meter reading and billing,this paper studies the gas meter reading recognition method based on the LeNet-5 model without changing the structure of gas meters.The main research contents are as follows:Aiming at the over-bright or over-dark gas meter images collected in complex environments,this paper proposes an image enhancement method based on brightness adaptive gamma transform.This method first adaptively adjusts the brightness of the image,and then enhances the image using the gamma transform method.The experimental results show that this method can effectively solve the problems of image overexposure and dim lighting.The target region of the gas meter image is located and segmented.The projection method is used to locate the reading area,and the combination of edge detection and mathematical morphology is used to locate the bar code area.The Hough transform method is used to correct the tilt of the gas meter image,and the vertical projection method is used to segment the located reading area into a single digital character.Aiming at the character characteristics of the gas meter reading area,a gas meter reading character recognition algorithm based on the improved LeNet-5 model is proposed,which mainly improves the activation function and pooling layer algorithm of the traditional LeNet-5 model.The experimental results show that compared with the traditional LeNet-5 model and BP neural network,the improved LeNet-5 model has higher recognition accuracy,better stability,and convergence.Building a gas meter image recognition system based on PyQt5,the system interface shows the recognition process and results of the gas meter image.The function and performance of the system are tested to verify the stability,feasibility,and anti-light interference ability of the system.The experimental results show that the recognition method designed in this paper has a strong anti-light interference ability and a good recognition effect.