基于改进YOLOv5s的拉挤板缺陷检测

doi:10.19936/j.cnki.2096-8000.20250528.017

摘要/Abstract

摘要： 为解决传统拉挤板缺陷检测方法中存在的检测精确度低、检测速度慢等问题,创建了玻璃纤维拉挤板缺陷数据集,提出了一种基于改进YOLOv5s的拉挤板缺陷检测模型。主要改进为:在特征提取网络部分,添加EvcBlock模块增强小目标特征提取能力,添加CBAM注意力机制提高重要特征的关注度;使用C3-Faster模块优化C3模块,实现了模型轻量化;在检测端引入具有形状损失的新型损失函数ShapeIoU,优化了预测框和真实框的拟合效果,提高了缺陷检测精确度。实验结果表明:改进后的YOLOv5s模型对比原YOLOv5s模型,mAP@0.5提升了3.6%,达到了88.7%,参数量降低了2.1%。改进模型检测速度为121.95 f/s,与YOLOv8s等五种模型相比综合性能更优,能够满足拉挤板缺陷检测的需求。

关键词: 拉挤板, YOLOv5s, 缺陷检测, EvcBlock, C3-faster, ShapeIoU, 复合材料

Abstract: In order to solve the problems of low detection precision and slow detection speed in traditional pultrusion plate defect detection methods, a defect data set of glass fiber pultrusion plate was created, and a defect detection model of glass fiber pultrusion plate based on improved YOLOv5s was proposed. The main improvements were as follows: in the feature extraction network part, EvcBlock module was added to enhance the feature extraction ability of small targets, and CBAM attention mechanism was added to improve the attention of important features; the lightweight model was realized by using C3-FASTER module to optimize the C3 module; a new type of loss function ShapeIoU with shape loss was introduced into the detection end, which optimized the fitting effect of the predicted frame and the real frame, and improved the precision of defect detection. The experimental results demonstrate that: in comparison to the original YOLOv5s model, the mAP@0.5 of the improved YOLOv5s model has increased by 3.6%, reaching 88.7%, while the number of parameters has been reduced by 2.1%. The detection speed of the improved model is 121.95 frames per second, and its overall performance is superior to 5 models such as YOLOv8s, which can meet the needs of pultrusion plate defect detection.

Key words: pultrusion plate, YOLOv5s, defect detection, EvcBlock, C3-faster, ShapeIoU, composites

中图分类号:

TB332

徐东亮, 赖九衡, 杨会兰. 基于改进YOLOv5s的拉挤板缺陷检测[J]. 复合材料科学与工程, 2025, 0(5): 132-141.

XU Dongliang, LAI Jiuheng, YANG Huilan. Defect detection of pultrusion plate based on improved YOLOv5s[J]. COMPOSITES SCIENCE AND ENGINEERING, 2025, 0(5): 132-141.

参考文献

[1] 薛忠民, 王占东, 尹证. 中国工业复合材料发展回顾与展望[J].复合材料科学与工程, 2021(6): 119-128.
[2] 陈博. 国内外复合材料工艺设备发展述评之五——拉挤成型[J].复合材料科学与工程, 2023(增刊1): 44-63.
[3] GIRSHICK R. Fast R-CNN[C]//IEEE.2015 IEEE International Conference on Computer Vision (ICCV). Santiago: IEEE, 2015: 1440-1448.
[4] REN S, HE K, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[5] HE K, GKIOXARI G, DOLLAR P, et al. Mask R-CNN[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2): 386-397.
[6] 向宽, 李松松, 栾明慧, 等. 基于改进Faster RCNN的铝材表面缺陷检测方法[J].仪器仪表学报, 2021, 42(1): 191-198.
[7] 陆尧, 薛林, 王云森, 等. 基于Cascade RCNN的热轧带钢表面缺陷检测[J].仪表技术与传感器, 2023(8): 101-106, 126.
[8] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//IEEE. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas: IEEE, 2016: 779-788.
[9] LIU W, ANGUELOV D, ERHAN D, et al. SSD: Single shot multibox detector[M]//LEIBE B, MATAS J, SEBE N, et al. Computer Vision-ECCV 2016: vol. 9905. Cham: Springer International Publishing, 2016: 21-37.
[10] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2): 318-327.
[11] TAN M, PANG R, LE Q V. EfficientDet: scalable and efficient object detection[C]//IEEE. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle: IEEE, 2020: 10778-10787.
[12] TIAN S, Li X, FANG X, et al. Surface defect detection method of wooden spoon based on improved YOLOv5 algorithm[J].BioResources, 2023, 18(4): 7713-7730.
[13] LI F, XIAO K, HU Z, et al. Fabric defect detection algorithm based on improved YOLOv5[J].The Visual Computer, 2024, 40(4): 2309-2324.
[14] 蒋博, 万毅, 谢显中. 改进YOLOv5s的轻量化钢材表面缺陷检测模型[J].计算机科学, 2023, 50(增刊2): 271-277.
[15] 渠逸, 汪诚, 余嘉博, 等. 基于YOLOv5的表面缺陷检测优化算法[J].空军工程大学学报, 2023, 24(5): 80-87.
[16] GHIASI G, LIN T Y, LE Q V. NAS-FPN: learning scalable feature pyramid architecture for object detection[C]//IEEE. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach: IEEE, 2019: 7029-7038.
[17] ZHOU L, RAO X, LI Y, et al. A lightweight object detection method in aerial images based on dense feature fusion path aggregation network[J].ISPRS International Journal of Geo-Information, 2022, 11(3): 189.
[18] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[M]//FERRARI V, HEBERT M, SMINCHISESCU C, et al. Computer Vision-ECCV 2018: vol. 11211. Cham: Springer International Publishing, 2018: 3-19.
[19] QUAN Y, ZHANG D, ZHANG L, et al. Centralized feature pyramid for object detection[J].IEEE Transactions on Image Processing, 2023, 32: 4341-4354.
[20] CHEN J, KAO S-H, HE H, et al. Run, don’t walk: chasing higher FLOPS for faster neural networks[C]//IEEE. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Vancouver: IEEE, 2023: 12021-12031.
[21] MA X, GUO F M, NIU W, et al. PCONV: the missing but desirable sparsity in DNN weight pruning for real-time execution on mobile devices[J].Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(4): 5117-5124.
[22] ZHANG H, ZHANG S. Shape-IoU: more accurate metric considering bounding box shape and scale[EB/OL].(2023-12-29)[2024-03-20].https://arxiv.org/abs/2312.17663.