针对风电叶片轻微裂纹难于检测的问题，提出一种基于改进 YOLOv5s模型的检测方法：通过在主干网络部分使用空洞空间金字塔池化（ASPP）代替空间金字塔池化（SPP）以适应不同大小和比例的目标，将注意力机制模块（squeeze and excitation，SE）插入主干网络中以增加网络对微小缺陷的敏感度，使用结构化交并比损失（SIoU-Loss）代替完全交并比损失（CIoU-Loss），以进一步提高新网络的准确性和训练速度。针对以上检测方法采用自建数据集进行对比实验，实验结果表明，改进 YOLOv5s 模型的平均精度均值（mAP）为 94.29%，与YOLOv5s模型相比提升了7.03个百分点，检测精度与其他的主流模型对比依然具有优势，检测速度为42.78 f/s。该方法在风电叶片内腔缺陷检测方面具有较好的使用性能和效果。
This paper proposes a detection method based on an improved YOLOv5s model for the problem of slight cracks in wind turbine blades that are difficult to detect. The method mainly includes three improvements. First, in the backbone network part, ASPP (atrous spatial pyramid pooling) is used instead of SPP (spatial pyramid pooling) to adapt to targets of different sizes and proportions. Second, SE (squeeze and excitation) attention modules are inserted into the backbone network to increase the network's sensitivity to small defects; and SIoU-Loss is used to replace the original CIoU-Loss to further improve the accuracy and training speed of the new network. Finally, a comparison experiment is conducted using a self-built dataset. Experimental results show that the mAP value of the improved YOLOv5s model is 94.29%, which is 7.03 percentage points higher than that of the YOLOv5s model, and its detection accuracy has advantages over other mainstream models. The detection speed is 42.78f/s. This method has good performance and effect in detecting defects inside wind turbine blades.
[1] 王新. 风电叶片典型缺陷的性能评估与模式识别[D]. 湘潭: 湘潭大学材料科学与工程学院, 2019.
[2] Chua L O. CNN: A vision of complexity[J]. International Journal of Bifurcation and Chaos, 1997, 7(10): 2219-2425.
[3] 陈威, 郭晓峰, 赵鹏飞, 等. 基于深度学习的风电叶片缺陷检测方法[J]. 电力系统自动化, 2020, 44(11): 1-8.
[4] Liu W, Anguelov D, Erhan D, et al. SSD: Single shot MultiBox detector[M]//Computer Vision-ECCV 2016. Cham: Springer International Publishing, 2016: 21-37.
[5] Jiang P Y, Ergu D J, Liu F Y, et al. A review of yolo algorithm developments[J]. Procedia Computer Science, 2022, 199: 1066-1073.
[6] Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. New York: ACM, 2014: 580-587.
[7] Ren S Q, He K M, 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.
[8] Cai Z W, Vasconcelos N. Cascade R-CNN: Delving into high quality object detection[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2018: 6154-6162.
[9] 李靖国, 张晓峰, 王晓东, 等. 基于声发射技术的复合材料风电叶片缺陷监测技术[J]. 机械工程学报, 2019, 55(13): 1-10.
[10] Movsessian A, García C D, Tcherniak D. An artificial neural network methodology for damage detection: Demonstration on an operating wind turbine blade[J]. Mechanical Systems and Signal Processing, 2021, 159: 107766.
[11] Yu Y J, Cao H, Yan X Y, et al. Defect identification of wind turbine blades based on defect semantic features with transfer feature extractor[J]. Neurocomputing, 2020, 376(C): 1-9.
[12] Wang D, Zhang Y F, Yang X Y. Image recognition of wind turbines blade surface defects based on maskRCNN[M]//Advanced Intelligent Technologies for Industry. Singapore: Springer Nature Singapore, 2022: 573-584.
[13] Lv L, Yao Z Y, Wang E M, et al. Efficient and accurate damage detector for wind turbine blade images[J]. IEEE Access, 2022, 10: 123378-123386.
[14] 郭晓峰, 赵鹏飞, 李靖国, 等. 基于红外热像仪的风电叶片缺陷检测方法[J]. 电力系统自动化, 2019, 43(7): 1-8.
[15] Regan T, Beale C, Inalpolat M. Wind turbine blade damage detection using supervised machine learning algorithms[J]. Journal of Vibration and Acoustics, 2017, 139(6): 061010.
[16] 刘磊, 张宇航, 马骏. 基于无人机的风电叶片巡检图像拼接技术[J]. 计算机应用与软件, 2020, 37(9): 1-5.
[17] 赵云龙, 李建军, 张宇航. 基于机器人的风电叶片缺陷检测系统设计与实现[J]. 计算机应用与软件, 2020, 37(10): 1-5.
[18] Qiu Z F, Wang S X, Zeng Z X, et al. Automatic visual defects inspection of wind turbine blades via YOLObased small object detection approach[J]. Journal of Electronic Imaging, 2019, 28(4): 043023.
[19] 杨洋洋, 张宇航. 基于云计算平台的风电叶片缺陷数据管理系统设计与实现[J]. 计算机应用与软件, 2020, 37(11): 1-5.
[20] He K M, Zhang X Y, Ren S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916.
[21] Hu J, Shen L, Sun G. Squeeze-and-excitation networks [C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2018: 7132-7141.
[22] Redmon J, Farhadi A. YOLOv3: An incremental improvement[J]. arXiv preprint arXiv:1804.02767, 2018.
[23] Bochkovskiy A, Wang C Y, Liao H Y M. YOLOv4: Optimal speed and accuracy of object detection[J]. arXiv preprint arXiv: 2004.10934, 2020.
