Research And Application Of MEC-based Detection Of Small Defects On Metal Surface

With the rapid development of the Industrial Internet of Things,the traditional Internet of Things technology collects data through sensors and transmits the data to the server for data analysis and logical processing when the amount of data is small,which can better store,forward and process.Key data in the production process.However,with the continuous growth of the Industrial Internet of Things and the exponential increase in data,the traditional server model is difficult to meet the realtime intelligent detection requirements in industrial applications.The traditional computing method is that the application program transmits the data to the cloud computing center,and enters the loop to wait for the data processing result.In the process of data processing waiting and transmission,a certain waiting delay needs to be added.And there are serious data security risks in the process of data transmission.The concept of Mobile Edge Computing(MEC)can effectively solve this problem,process and analyze data close to the data producer,reduce service requests to cloud servers,and shorten the delay of the overall system.The shortening of data transmission links enables edge computing to respond quickly and efficiently to business needs on the data generation side.Data storage,processing,and analysis at the edge can also improve user privacy protection.This in turn reduces the reliance on uncertain network links.Combine edge computing with traditional industries to improve the industrial Io T industry in the new era.In order to solve the problems faced in the process of industrial transformation and upgrading,edge computing can coordinate with the data processing mode centered on the cloud computing model,expand the data processing capabilities of cloud computing at the edge network information production end,and improve industrial transformation and upgrading.This paper firstly introduces the background and significance of the selected topic research,and introduces the research status of edge computing and deep learning related technologies at home and abroad.Secondly,it focuses on the defect detection algorithm designed in this paper.Swin Transformer is used as the feature extraction network,and the cascaded multi-threshold detection structure is used as the output layer of the deep learning surface defect detection algorithm.The Transformer structure is applied to the field of strip steel surface defect detection,which can achieve a more accurate detection effect than the deep learning target detection algorithm based solely on convolutional networks.Compared with deep learning algorithms such as Yolov3,Yolo F,Deform Detr,SSD512 and SSD Lit,it achieves better defect detection accuracy in industrial hot-rolled strip defect data.In the crack(Crazing,Cr),inclusion(Inclusion,In),plaque(Patches,Pa),pitting(Pitted Surface,PS),pressed into the iron oxide scale(Rolled-in Scale,RS),and scratches(Scratches,In the detection of surface defects such as Sc),the training speed and detection accuracy are significantly improved,and the missed detection rate is significantly reduced.The algorithm can meet the requirements of industrial Internet of Things applications for defect detection accuracy.Then,it focuses on the expansion and application of edge computing,Internet of Things,and deep learning algorithms in industrial hot-rolled strip application scenarios.Through the deployment of the distributed microservice framework Edge X Foundry in the container,the industrial edge detection system environment is provided for the design and application of the detection algorithm for small defects on the surface of hot rolled strip.The edge-side video server is built as the algorithm inference center,where the video images obtained from the bottom layer are used to implement the defect detection and inference function,and the inference results are sent to the bottom layer device side for timely response.The algorithmic inference module is deployed on the edge video server side.The video server transmits video data to the upper-layer cloud server for algorithmic model training.The trained algorithmic inference model is deployed to the edge-side video server for actual inference execution.The research and application of the detection of tiny defects on metal surfaces based on edge computing is a combination of edge computing and artificial intelligence algorithms in actual industrial application scenarios,as well as an exploration to realize the concept of edge artificial intelligence.It is of great research significance to integrate artificial intelligence algorithms to realize the task of small defect detection while protecting the problem.