基于深度学习的工业纸张图像批量化检测处理技术研究
Research on Batch Detection and Processing Technology of Industrial Paper Images Based on Deep Learning
周小娟 1商娟叶1
作者信息
- 1. 西安外事学院,陕西 西安,710077
- 折叠
摘要
为实现针对纸张缺陷的自动化批量检测,提高造纸企业工业纸张生产的合格率水平.通过融合空间注意力机制对EfficientNet算法中的卷积神经网络进行优化,进而形成了一套SAEfficientNet算法,并通过该算法实现了对工业纸张图像的批量化检测,大批量识别出了纸张图像中的裂缝缺陷.为验证该方法的应用性能,挑选出800 张纸张缺陷图像,并通过纸张缺陷图像数据对算法模型进行检测与训练.研究结果表明:SAEfficientNet算法对纸张缺陷识别的准确率高达 98.19%,具有一定的应用价值.
Abstract
In order to achieve automated batch detection of paper defects and improve the qualification rate of industrial paper production in papermaking enterprises.This study optimized the convolutional neural network in EfficientNet algorithm by integrating spatial attention mechanism,and formed a set of SAEfficientNet algorithm.Through this algorithm,batch detection of industrial paper images was achieved,and cracks and defects in paper images were identified in large quantities.To verify the application performance of this method.This study selected 800 paper defect images and used paper defect image data to detect and train the algorithm model.Through experimental research,it was found that the SAEfficientNet algorithm has a high accuracy rate of 98.19%for paper defect recognition,which has certain application value.
关键词
复合模型/SAEfficientNet算法/空间注意力机制/数字图像识别Key words
composite model/SAEfficientNet algorithm/spatial attention mechanism/digital image recognition引用本文复制引用
出版年
2024
NETL
NSTL
万方数据