Hardware Image Recognition And Retrieval System Based On Edge Computing

In today’s era,digital technology and digital economy are the opportunities for the world’s technological revolution and industrial transformation,and are the key areas of a new round of international competition.China is a big bag production country since there are four PVC bag production and manufacturing bases,formed by Guangdong Huadu,Fujian Quanzhou,Zhejiang Pinghu and Hebei Baigou.In view of the problem that the current luggage hardware enterprises are mostly small and medium-sized enterprises,and the digitalization level of the production process of the enterprise is low,the subject is committed to solving the service and production efficiency caused by the wide variety of hardware and the manual selection of hardware in the process of luggage production and sales.Due to the low productivity and high operating costs,a lightweight hardware image classification model based on improved Shuffle Net V2 is proposed,and a hardware image recognition and retrieval system is developed combined with edge computing to achieve automatic and efficiently identify and retrieve the required hardware parts,effectively improve service and production efficiency,and reduce operating costs.The main research contents and phased results of the thesis are as follows:(1)Aiming at the problem of how to reduce the complexity of hardware image classification model and achieve higher classification effect in practical application scenarios,a hardware image classification model(named ECA-Shuffle Net)is proposed to combine the Shuffle Net V2 network and the ECA-Net attention mechanism.The effectiveness of the proposed hardware image classification model is verified by comparative experiments on real hardware image data sets,and it is applied to the hardware image recognition and retrieval system.(2)Aiming at the problem of how to reduce computing resource consumption and image transmission delay to improve the real-time performance of hardware image classification model in practical application scenarios,a system architecture based on edge computing and deep learning is proposed.In this architecture,the terminal is responsible for collecting and sending hardware images,the cloud device is responsible for training the convolutional neural network model,and the edge device is responsible for the recognition and retrieval of hardware images.The proposed architecture is compared with the hardware image recognition and retrieval architecture based on traditional cloud computing,and the feasibility and effectiveness of the architecture are verified.(3)Combining edge computing architecture and improved network model to develop hardware image recognition and retrieval system,the software engineering method is used to carry out the feasibility analysis,demand analysis,overall design and database design for hardware image recognition and retrieval system,and the design and implementation of the main functional modules of the system are described in detail,and finally a unified software testing is carried out,the test results are in line with expectations.