Treffer: 风机叶片表面缺陷的小样本检测方法.

In response to the challenge posed by data-driven deep learning methods, which require a substantial amount of annotated data, this study addresses the difficulty of obtaining such data in practical applications. Specifically, we propose a few-shot defect detection method for wind turbine blade surfaces. Currently, research on two-stage object detection models, particularly Faster R-CNN, is a hot topic in few-shot learning. Building upon this foundation, we introduce a sample feature aggregation (SFA) approach. Our method leverages a variational autoencoder to learn the latent distribution of support set samples and subsequently samples category attention vectors from it. This strategy mitigates the impact of poorly defined features in individual samples on the category attention vectors. Remarkably, our proposed method achieves effective defect detection on wind turbine blade surface defects using only five images from a self-constructed dataset. Experimental results demonstrate its superiority over other comparative methods in the context of few-shot defect detection. [ABSTRACT FROM AUTHOR]

针对数据驱动的深度学习方法需要大量标注数据而在实际应用中获取数据困难的问题, 提出风机叶 片表面缺陷的小样本检测方法。 目前在小样本学习中, 基于两阶段目标检测模型 Faster R -CNN 的研究是一个热 点。 在此基础上, 提出了一个采样特征聚合 (sample feature aggregation, SFA) 的小样本缺陷检测方法。 该方法引入 变分自编码器学习支持集样本的潜在分布, 并从中采样出类别注意力向量, 避免个别特征不明显的样本对类别注 意力向量的影响。 在自主构建的风机叶片表面缺陷 (wind turbine blade surface defect, WSD) 数据集上, 所提出的方 法仅用 5 张图片就能实现对叶片表面缺陷的检测。 实验结果表明, 所提方法在小样本缺陷检测问题中优于其他对 比方法。 [ABSTRACT FROM AUTHOR]

Copyright of Journal of Harbin University of Science & Technology is the property of Journal of Harbin University of Science & Technology and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)