%0 Journal Article
%T 基于宽度学习的图像重建与缺陷检测方法<br>Image Reconstruction and Defect Detection Method Based on Broad Learning System
%A 郑泽华
%J Computer Science and Application
%P 579-586
%@ 2161-881X
%D 2023
%I Hans Publishing
%R 10.12677/CSA.2023.133057
%X <span style=\"font-size:14px;\">针对在训练中缺乏足量的缺陷样本会导致有监督的深度学习检测方法难以展开，并且常见的无监督缺陷检测方法面临着精度不足，效率低等问题。我们提出了一种基于宽度学习的图像重建与缺陷检测方法，该方法基于宽度学习系统改进，使其具有重构完整图像的能力，通过将测试图像与重构图像进行对比，确定测试图像中的缺陷。实验结果表明，该方法具有较好的重构精度，在BTAD数据集具有更好的效果。该模型的平均训练时间在325 s左右，运行速度达到60帧/s，优于常见的深度学习检测方法，基本能够满足实时检测的需求。</span><br />
<span style=\"font-size:14px;\">The lack of sufficient defect samples in training will make it difficult to develop supervised deep learning detection methods, and common unsupervised defect detection methods face problems such as insufficient accuracy and low efficiency. We propose an image reconstruction and defect detection method based on broad learning, which improves the broad learning system to make it achieves the ability to reconstruct the complete image, and determine the defects in the test image by comparing it with the reconstructed image. The experimental results show that the proposed method has good reconstruction accuracy and better effect in BTAD dataset. The average training time of the model is about 325 s, and the running speed reaches 60 frames per second, which is better than common deep learning detection methods and can basically meet the needs of real-time detection.</span>
%K 缺陷检测，宽度学习系统，图像重建，无监督学习<br>Defect Detection
%K Broad Learning System
%K Image Reconstruction
%K Unsupervised Learning
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=63531