基于深度学习和生成对抗网络的发动机缸体表面缺陷检测方法

doi:10.3901/JME.2025.02.046

摘要/Abstract

摘要： 针对某企业发动机缸体表面缺陷检测的工业视觉应用中，由于缺陷样本数据集较少导致的识别准确率低，检测速度慢的问题，设计了一套基于深度学习的机器视觉表面缺陷检测系统。采用生成对抗网络pix2pix模型增强缺陷图片数据集，结合灰度增强、带通滤波等传统图像预处理方法突出表面缺陷特征，利用YOLOv5深度学习识别网络进行缺陷识别算法开发，并根据上述框架搭建了缸体表面缺陷检测软硬件系统。将所搭建的检测系统在企业生产线进行应用验证，结果显示，缺陷总体识别准确率高达98.4%，单张图片识别耗时小于0.5 s。所提方法解决了原有小样本数据集难以支撑深度学习训练的难题，提高了缸体表面缺陷的识别精度和识别效率，在工业检测的应用中具有巨大的潜力。

关键词: 发动机缸体, 缺陷检测, YOLOv5, pix2pix

Abstract: In the industrial vision application for engine cylinder surface defect detection, the lack of defect datasets leads to low detection accuracy and efficiency. An engine cylinder surface defect detection algorithm based on the YOLOv5 network and Pix2Pix model is proposed. The generative adversarial network pix2pix model is used for dataset enhancement. The image pre-processing methods, such as the grayscale enhancement method and band-pass filtering method, are used to highlight the surface defect features. The defect detection algorithm is developed based on the YOLOv5 model. Then a cylinder surface defect detection hardware and software system is established using the above framework. Experimental validation is conducted on the production line in the industry. Results show that the detection accuracy of the proposed method is 98.4%. Single image detection takes less than 0.5 s. This research has the capacity to deal with the lack of datasets in deep learning training, which improves the detection accuracy and efficiency of engine cylinder surface defect detection. The proposed method shows great potential for industrial inspection applications.

Key words: engine cylinder, defect detection, YOLOv5, pix2pix

中图分类号:

TP391

曾治霖, 瞿昊, 杜正春. 基于深度学习和生成对抗网络的发动机缸体表面缺陷检测方法[J]. 机械工程学报, 2025, 61(2): 46-55.

ZENG Zhilin, QU Hao, DU Zhengchun. An Engine Cylinder Surface Defect Detection Algorithm Based on the YOLOv5 Network and Pix2Pix Model[J]. Journal of Mechanical Engineering, 2025, 61(2): 46-55.

参考文献

[1] 殷苏民，鲍红力，吉彬斌，等. 基于小区域模板匹配的发动机缸体缺陷检测[J]. 传感器与微系统，2012，31(6)：143-145. YIN Sumin，BAO Hongli，JI Binbin，et al. Engine block defect detection based on small area template matching[J]. Sensors and Micro Systems，2012，31(6)：143-145.
[2] 大连亚明汽车部件股份有限公司. 一种基于机器视觉的发动机缸体外观缺陷检测算法：CN202110401936. 5[P]. 2021-07-23. Dalian Yamin Auto Parts Co. A machine vision-based engine block appearance defect detection algorithm：CN202110401936.5[P]. 2021-07-23.
[3] KOU Q，WANG L，LI D，et al. A method of electricity meter lcd screen defect detecting based on convolutional neural network [C]//2021 China Automation Congress (CAC)，2021：7553-7558.
[4] ZHANG Z. Detecting LED chip surface defects with modified faster R-CNN [C]//202118th International SoC Design Conference (ISOCC)，2021：268-269.
[5] KIM S，JEON T，YEO J，et al. Camera module defect detection using Gabor filter and convolutional neural network [C]//2020 International Conference on Electronics，Information，and Communication (ICEIC)，2020：1-4.
[6] 李建乐，张佳奇，黄念，等. 基于深度学习的分布式光纤损伤识别方法[J]. 机械工程学报，2022，58(8)：88-95. LI Jianle，ZHANG Jiaqi，HUANG Nian，et al. A distributed optical fiber damage identification method based on deep learning [J]. Journal of Mechanical Engineering，2022，58(8)：88-95.
[7] URAON P K，VERMA A，BADHOLIA A. Steel sheet defect detection using feature pyramid network and RESNET [C]//2022 International Conference on Edge Computing and Applications (ICECAA)，2022： 1543-1550.
[8] 姜洪权，贺帅，高建民，等. 一种改进卷积神经网络模型的焊缝缺陷识别方法[J]. 机械工程学报，2020，56(8)：235-242. JIANG Hongquan，HE Shuai，GAO Jianmin，et al. An improved convolutional neural network model for weld defect identification [J]. Journal of Mechanical Engineering，2020，56(8)：235-242.
[9] YAAN A C. Investigation of guided filtering for image processing applications [C]//202230th Signal Processing and Communications Applications Conference (SIU)，2022：1-4.
[10] ZHAO H. Multiscale vascular enhancement filter applied to in vivo morphologic and functional photoacoustic imaging of rat ocular vasculature [J]. IEEE Photonics Journal，2019，11(6)：1-12.
[11] UERKAN，THANH D N H，HIEU L M，et al. An iterative mean filter for image denoising [J]. IEEE Access，2019，7：167847-167859.
[12] GOODFELLOW I J，POUGET-ABADIE J. Generative adversarial networks [J]. Commun. ACM，2020，63(11)：139-144.
[13] YANG S，BAI H. Research on the application of double conditional general advantage nets (DCGAN) to generate simple background image with specified hue[C]//Journal of Physics：Conference Series. IOP Publishing，2020，1486(2)：022044.
[14] LEE E H，JEON M，JANG H，et al. Study on the training effectiveness of deep learning with synthesized underwater sonar image using Pix2Pix and FCN [C]//2020 IEEE/OES Autonomous Underwater Vehicles Symposium (AUV)，2020：1-3.
[15] HE J，WANG C，JIANG D，et al. CycleGAN with an improved loss function for cell detection using partly labeled images [J]. IEEE Journal of Biomedical and Health Informatics，2020，24(9)：2473-2480.
[16] YANG Y，ZHANG J，LI Y，et al. GAN-based method with small training samples for defect detection using magneto-optical image [C]//2021 CAA Symposium on Fault Detection，Supervision，and Safety for Technical Processes (SAFEPROCESS)，2021：1-4.
[17] YANG J，LIU J，XIE J，et al. Conditional GAN and 2-D CNN for bearing fault diagnosis with small samples [J]. IEEE Transactions on Instrumentation and Measurement，2021，70：1-12.
[18] JOSEPH R，SANTOSH K D，ROSS B G，et al. You only look once：Unified，real-time object detection [C/CD]//The 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas，USA，2016.
[19] ZHANG Jiaqiao，KANG Xin，NI Hongjun，et al. Surface defect detection of steel strips based on classification priority YOLOv3-dense network [J]. Ironmaking & Steelmaking，2021，48(5)：1-12.