改进的YOLOv5风机叶片缺陷检测方法
The Improved YOLOv5 Wind Turbine Blade Defect Detection Method
DOI: 10.12677/MOS.2023.124329, PDF,  被引量    国家自然科学基金支持
作者: 高文俊*, 张海峰：上海工程技术大学机械与汽车工程学院，上海
关键词: YOLOv5；风机叶片；注意力机制；多尺度；YOLOv5； Wind Turbine Blade； Attention Module； Multi Scale
摘要: 为了解决风力发电机叶片传统检测的耗时长，效率低，精度低等问题。为此，本文提出一种改进的YOLOv5风机叶片缺陷检测算法。首先，针对图像中复杂背景和图片模糊等影响因素，本算法采用限制对比度自适应直方图均衡化的方法，加强叶片缺陷特征，减弱这些因素对数据集的影响。其次，在主干网络中引入极化注意力机制，以增强网络对缺陷特征的敏感性，进一步提高模型检测能力。最后，在主干网络末端加入金字塔池化模块，以扩大网络不同尺度的感受能力，增强对多尺度目标的识别能力。实验结果表明，在自制风机叶片数据集中，改进的YOLOv5算法相比原有算法，召回率上升了13.7个百分点，准确率也增长了16.3个百分点，平均精度均值提高了16.5个百分点。因此，该算法可以更好地应用于风机叶片缺陷检测场景，为增强风力发电机组的操作效率和稳定性提供了有力支持。
Abstract: To address the issues of long detection times, low efficiency, and low accuracy in traditional detec-tion methods for wind turbine blades, this paper proposes an improved YOLOv5 algorithm for de-tecting blade defects. Firstly, to enhance the blade defect features and reduce the influence of com-plex backgrounds and blurry images, this algorithm uses a contrast-limited adaptive histogram equalization method. Secondly, a polarized attention mechanism is introduced into the backbone network to enhance the network’s sensitivity to defect features and further improve the model’s detection capabilities. Finally, a pyramid pooling module is added to the end of the backbone net-work to expand the network’s perception capabilities at different scales and enhance its recognition capabilities for multi-scale targets. Experimental results show that, compared with the original al-gorithm, the improved YOLOv5 algorithm increases the recall rate by 13.7 percentage points, the accuracy rate by 16.3 percentage points, and the average precision by 16.5 percentage points in the self-made wind turbine blade dataset. Therefore, this algorithm can be better applied to the wind turbine blade defect detection scenario, providing strong support for enhancing the efficiency and stability of wind power generation units.
文章引用：高文俊, 张海峰. 改进的YOLOv5风机叶片缺陷检测方法[J]. 建模与仿真, 2023, 12(4): 3574-3586. https://doi.org/10.12677/MOS.2023.124329

参考文献

[1] 曹庆才, 吴立东, 张路娜, 张舒翔, 徐志轩. 基于机器视觉的风机叶片缺陷检测[J]. 电工技术, 2021(22): 74-76+155. [Google Scholar] [CrossRef] 
[2] 严海领. 基于无人机的不同颜色风叶表面缺陷检测算法研究[J]. 现代信息科技, 2021, 5(15): 169-172+175. [Google Scholar] [CrossRef] 
[3] 邹洁, 何超, 朱永凯. 基于FBG和小波包能量谱的风机叶片无损检测[J]. 电子测量技术, 2015, 38(3): 133-138. [Google Scholar] [CrossRef] 
[4] Pizer, S.M., Amburn, E.P., Austin, J.D., et al. (1987) Adaptive Histogram Equalization and Its Variations. Computer Vision, Graphics, and Image Processing, 39, 355-368. [Google Scholar] [CrossRef] 
[5] Liu, H., Liu, F., et al. (2021) Polarized Self-Attention: Towards High-Quality Pixel-Wise Regression. Neurocomputing, 506, 158-167. [Google Scholar] [CrossRef] 
[6] Zhao, H., Shi, J., Qi, X., et al. (2017) Pyramid Scene Parsing Network. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, 21-26 July 2017, 2881-2890. [Google Scholar] [CrossRef] 
[7] Bochkovskiy, A., Wang, C.Y. and Liao, H.Y.M. (2020) Yolov4: Optimal Speed and Accuracy of Object Detection.
https://arxiv.org/abs/2004.10934
[8] Wang, C.Y., Liao, H.Y.M., Wu, Y.H., et al. (2020) CSPNet: A New Backbone That Can Enhance Learning Capability of CNN. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, Seattle, 14-19 June 2020, 390-391. [Google Scholar] [CrossRef] 
[9] Lin, T.Y., Dollár, P., Girshick, R., et al. (2017) Feature Pyramid Networks for Object Detection. Proceedings of the IEEE Confer-ence on Computer Vision and Pattern Recognition, Honolulu, 21-26 July 2017, 2117-2125. [Google Scholar] [CrossRef] 
[10] Liu, S., Qi, L., Qin, H., et al. (2018) Path Aggregation Network for Instance Segmentation. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, 18-22 June 2018, 8759-8768. [Google Scholar] [CrossRef] 
[11] He, Y., Zhu, C., Wang, J., et al. (2019) Bounding Box Regression with Uncertainty for Accurate Object Detection. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni-tion, Long Beach, 15-20 June 2019, 2888-2897. [Google Scholar] [CrossRef] 
[12] Deng, J., Zhang, S., Zhang, J., et al. (2020) Loss Function and Application Research in Supervised Learning. Big Data Research, 6, 60-80.
[13] Redmon, J. and Farhadi, A. (2017) YOLO9000: Better, Faster, Stronger. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, 21-26 July 2017, 7263-7271. [Google Scholar] [CrossRef] 
[14] Redmon, J. and Farhadi, A. (2018) YOLOv3: An Incremental Improvement. [Google Scholar] [CrossRef] 
[15] 陈科圻, 朱志亮, 邓小明, 等. 多尺度目标检测的深度学习研究综述[J]. 软件学报, 2021, 32(4): 1201-1227.
[16] Hu, J., Shen, L. and Sun, G. (2018) Squeeze-and-Excitation Networks. Pro-ceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, 18-22 June 2018, 7132-7141. [Google Scholar] [CrossRef] 
[17] Hu, J., Shen, L., Albanie, S., et al. (2018) Gather-Excite: Exploiting Fea-ture Context in Convolutional Neural Networks. 32nd Conference on Neural Information Processing Systems (NIPS 2018), Montréal, 3-8 December 2018, 31-41.
[18] Woo, S., Park, J., Lee, J.Y., et al. (2018) CBAM: Convolutional Block Attention Module. Proceedings of the European Conference on Computer Vision (ECCV), Munich, 8-14 September 2018, 3-19. [Google Scholar] [CrossRef] 

为你推荐