Application of Diagram Image Classification Based on PCBMER. Issue 2 (June 2021)

• Record Type: Journal Article
• Authors: Wang, Ruixiang
• Is Part Of: Journal of physics. Volume 1952:Issue 2(2021)
• Publication Date: 2021-06
• Subjects: Component;chart classification -- Convolution neural network -- PCBMER architecture -- Deep learning; Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/1952/2/022002
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: Nowadays, in order to express data more clearly and accurately, data visualization has become a wave of the times. In this context, due to the different types and quantities of data stored in computer systems, different forms of data visualization can be seen in web pages, demonstration slides and papers. However, in many cases, we can't directly obtain the underlying data, so the classification of chart images becomes very important. Only with the classification technology of chart images can we understand it more intuitively and deal with it more concretely, such as generating new charts according to the extracted data. Using depth learning technology to classify chart images is now a very popular way. In the field of image classification, convolution neural network (CNN) is the most widely used technology in depth learning, not only because of its high accuracy, but also because it encapsulates the process of feature extraction. At the same time, using the software architecture of PCBMER framework to design application programs can maintain high stability of the system. From the experimental results, this new technology is feasible, and the architecture can be further optimized to make the results more accurate.