Metallographic Grade Recognition and Data Analysis Based on 6G Industrial Internet

With the development of modern image processing technology, the introduction of image processing and analysis technology into metallographic microstructure analysis has become a key research direction. Through the effective analysis and measurement of metal microstructure and structure, and even the prediction, application and design of material properties, the quantitative analysis of gold phase can be realized. Using machine learning algorithm can partially replace the task of manual rating, and make up for the shortcomings of high intensity and poor repeatability of human detection fees. In this paper, in order to solve the problems such as high requirements for identifying grain size and large amount of retained information in metallographic atlas, we use up sampling and convolution operations to restore and preprocess the image, build a convolution neural network, predict and segment each pixel separately, and determine the maximum pooling and average pooling proportion in the metallographic atlas sampling process, so as to retain more feature information in the process of bottom feature extraction. Aiming at the problems of resource isolation, high transmission delay, low data fusion efficiency and low analysis accuracy of metallographic atlas in the process of industrial Internet data analysis, a federated learning-driven industrial Internet based on 6G enabled metallographic atlas data collaboration, fusion and analysis method is proposed. Based on the cross layer aggregation high-precision data analysis method of integrated learning, the aggregation strategy of synchronization and hierarchical link between layers within the layer is constructed to realize the metallographic map recognition and data analysis system, and improve the accuracy of intelligent decision-making of industrial Internet business based on 6G.