New Data from Shandong University Illuminate Findings in Networks (Unsupervised Cnn-based Dic Method for 2d Displacement Measurement)

• 全文链接
• 请求原文

关键词：

Neura l Networks;  Supervised Learning;  Networks;  Emerging Technologies;  Asia;  Jinan;  Convolutional Network;  Shandong University;  People's Republic of China;  Machine Learning; 

期刊名称：

Network Daily News

i s s n：

年卷期：

2024 年 Mar.6 期

页   码：

46-47

页   码：

摘   要：

Data detailed on Networks have been presented. According to news reporting from Jinan, People's Republic of China, by NewsRx journalists, research stated, “Digital image correlation (DIC) is a widely used technique for noncontact mea surement of deformation. However, traditional DIC methods face challenges in bal ancing calculation efficiency and the quantity of seed points.” Financial support for this research came from National Natural Science Foundatio n of China (NSFC). The news correspondents obtained a quote from the research from Shandong Univers ity, “Deep learning approaches, particularly supervised learning methods, have s hown promise in improving DIC efficiency. However, these methods require high-qu ality training data, which can be time-consuming to generate ground truth annota tions. To address these challenges, we propose an unsupervised convolutional neu ral network (CNN) based DIC method for 2D displacement measurement. Our approach leverages an encoderdecoder architecture with multi-level feature extraction, a dual-path correlation block, and an attention block to extract informative feat ures from speckle images with varying characteristics. We utilize a speckle imag e warp model to transform the deformed speckle image to the predicted reference speckle image based on the predicted 2D displacement map. The unsupervised train ing is achieved by comparing the predicted and original reference speckle images . To optimize the network's parameters, we employ a composite loss function that takes into account both the Mean Squared Error (MSE) and Pearson correlation co efficient. By using unsupervised convolutional neural network (CNN) based DIC me thod, we eliminate the need for extensive training data annotation, which is a t ime-consuming process in supervised learning DIC methods. We have conducted seve ral experiments to demonstrate the validity and robustness of our proposed metho d. The results show a significant reduction in Mean Absolute Error (MAE) and Roo t Mean Squared Error (RMSE) compared to a method proposed by Zhao et al. This in dicates that our unsupervised CNN-based DIC approach can achieve accuracy compar able to supervised CNN-based DIC methods.