Research On The Identification Method Of Authenticity Of Two-dimensional Code Image For Elevator Maintenance Management

With the rapid development of China’s economy,national strength becomes stronger and stronger,and people’s living standards are constantly improved.Nationwide infrastructure is constantly being upgraded too.As one of the most important infrastructures,the number of elevators is also increasing.China has now become the country with the most elevators in the world.However,with such a large number of elevators,how to ensure their safe and stable operation is even more important.In order to ensure the quality and safety of the elevator,we must implement maintenance work on the elevator regularly and efficiently.This article first proposes the method of image recognition to help related companies and governments improve elevator maintenance management.The main work of this article includes:(1)Establish "Image Dataset On Elevator Maintenance Management".This paper uses deep learning to study the image recognition method of elevator maintenance management.Deep learning requires a lot of training data.This article first determines the number and meaning of the category labels in the dataset according to the task requirements,and then cleans the image data downloaded from the original database through the "multi-scale template matching" method.Finally,the classification and label of the cleaned data are carried out according to the judgment standards formulated in this article." Image Dataset On Elevator Maintenance Management" will provide training and test data for all experiments in this article.(2)A deep neural network based on the Inception architecture is constructed and used for elevator maintenance management image recognition.The image dataset on elevator maintenance and management constructed in this paper has the characteristics of less data volume and high ambiguity between classes.The Inception module can greatly reduce the amount of network parameters while maintaining strong expressiveness.This paper constructs a convolutional neural network based on Inception structure and applies it to the elevator maintenance management image recognition.Fewer parameters can prevent the network from overfitting in the face of small dataset,and Inception’s sparse structure can better adapt to different forms of features.(3)A Spatial-SE attention module is designed based on the spatial attention mechanism.Different from general image recognition tasks,the key point of elevator maintenance management image recognition is the recognition of the image background.Therefore,this paper introduces the attention mechanism in deep learning.It is hoped that the neural network will give more attention to background information of image when it recognizes it,thereby improving the recognition accuracy of the network.This paper draws on the ideas of SE-Net,designs the Spatial-SE spatial attention module,and constructs Spatial-SE-Inception network based on spatial attention and Mixed-SEInception network based on mixed attention for elevator maintenance m anage image recognition.