吉林大学学报(工学版) ›› 2024, Vol. 54 ›› Issue (5): 1205-1213.doi: 10.13229/j.cnki.jdxbgxb.20231081

• 车辆工程·机械工程 • 上一篇    下一篇

汽车漆面缺陷高精度检测系统

陆玉凯1,2(),袁帅科3,熊树生1,4(),朱绍鹏1,张宁3   

  1. 1.浙江大学 动力机械及车辆工程研究所，杭州 310014
    2.浙江吉利汽车有限公司，杭州 310051
    3.燕山大学 机械工程学院，河北 秦皇岛 066004
    4.龙泉产业创新研究院，浙江 龙泉 323700
  • 收稿日期:2023-10-10 出版日期:2024-05-01 发布日期:2024-06-11
  • 通讯作者: 熊树生 E-mail:yukai.lu@geely.com;xiongss@zju.edu.cn
  • 作者简介:陆玉凯（1980-），男，高级工程师，博士研究生. 研究方向：机器视觉，缺陷检测，汽车质量控制.E-mail： yukai.lu@geely.com
  • 基金资助:
    工信部重点领域及特定场景工业互联网平台应用项目(TC200802D);工信部5G+工业互联网”高质量网络和公共服务平台-离散行业高质量网络项目(TC200A00N)

High precision detection system for automotive paint defects

Yu-kai LU1,2(),Shuai-ke YUAN3,Shu-sheng XIONG1,4(),Shao-peng ZHU1,Ning ZHANG3   

  1. 1.Power Machinery & Vehicular Engineering Institute，Zhejiang University，Hangzhou 310014，China
    2.Zhejiang Geely Automobile Co. ，Ltd. ，Hangzhou 310051，China
    3.College of Mechanical Engineering，Yanshan University，Qinhuangdao 066004，China
    4.Longquan Industrial Innovation Research Institute，Longquan 323700，China
  • Received:2023-10-10 Online:2024-05-01 Published:2024-06-11
  • Contact: Shu-sheng XIONG E-mail:yukai.lu@geely.com;xiongss@zju.edu.cn

RICH HTML

 2   

PDF (PC)

107

摘要：

汽车涂装过程中产生的漆面缺陷影响着整车外观质量，针对人工检测存在漏检、低效以及传统检测方案的高实施成本等问题，提出了一种基于改进YOLOv7算法的汽车漆面缺陷检测系统。构建了汽车漆面缺陷数据集，共有4023张图像，其中包含5种常见汽车漆面缺陷；针对YOLOv7算法在微小缺陷上检测精度不足的问题，在原网络中引入了GAM注意力机制和SPPFCSPC模块，用于提高算法对微小缺陷特征的提取能力，同时采用改进的ELAN模块对网络结构进行改进，减少网络过深造成的小目标信息丢失问题，保证在减轻网络模型的同时提高网络对微小特征的识别精度；实验结果表明：本文方法大幅提升了对微小漆面缺陷的检测性能，缺陷的平均检测精度达到了88.9%，与多种算法相比检测精度最高。

关键词: 车辆工程, 汽车漆面, 缺陷检测, 深度学习

Abstract:

The paint defects that exist during the automotive painting process affect the overall appearance quality of the car. In response to the problems of missed inspection，low efficiency，and high implementation cost of traditional inspection schemes in manual inspection，a paint defect detection method based on the improved YOLOv7 algorithm is proposed. A dataset of automotive paint defects was constructed，consisting of 4023 images，including 5 types of automotive paint defects； In response to the problem of insufficient detection accuracy of YOLOv7 algorithm on small defects，GAM attention mechanism and SPPFCSPC module were introduced into the original network to improve the algorithm's ability to extract small defect features. At the same time，an improved ELAN module was used to improve the network structure to reduce the problem of small target information loss caused by deep network，ensuring that the network model is reduced while improving the recognition accuracy of small features； Based on the constructed dataset，the defect detection performance of different algorithms was tested and the effectiveness of the module was verified. The experimental results show that this method significantly improves the detection ability of small defects on paint surfaces，with an average detection accuracy of 88.9%，which is the highest detection accuracy compared to various algorithms.

Key words: vehicle engineering, automotive paint surface, defect detection, deep learning

中图分类号: 

  • TP183

图1

缺陷图像采集平台"

表1

设备配置参数"

名称配置
工业相机Basler acA5472-5gc
相机分辨率5472×3648
像元尺寸2.4 μm×2.4 μm
镜头大恒HN-P-1624-25M-C1.2/1
工作距离450 mm
工业光源开孔面光源

图2

YOLOv7 目标检测算法网络结构图"

图3

ELAN模块"

图4

SPPCSPC模块"

图5

GAM注意力机制"

图6

SPPFCSPC模块"

图7

ELAN-S模块"

表2

实验环境参数"

参数配置
CPUIntel Xeon Gold 6271C @2.6 GHz×20
GPUTesla V100 SXM2；32 G
CUDA10.2
系统Ubuntu 18.04.6LTS
语言Python 3.8

图8

不同数量ELAN-S的精度对比"

表3

改进YOLOv7算法的消融实验"

GAMSPPFCSPCELAN-SmAP@0.5
///0.719
√//0.811
√√/0.830
√√√0.889

图9

不同模型的热力图"

图10

不同缺陷的检测结果"

表4

不同检测方法的对比结果"

算法mAP@0.5mAP@0.5∶0.95Time/ms
Faster-RCNN0.8090.424119.1
YOLOv40.7060.34924.4
YOLOR0.8010.42324.3
YOLOv70.7190.36715.0
改进YOLOv70.8890.47116.3
1 Yu J B, Lu X L. Wafer map defect detection and recognition using joint local and nonlocal linear discriminant analysis[J]. IEEE Transactions on Semiconductor Manufacturing,2016,29(1): 33-43.
2 刘培勇,董洁,谢罗峰,等. 基于多支路卷积神经网络的磁瓦表面缺陷检测算法[J]. 吉林大学学报: 工学版,2023,53(5): 1449-1457.
Liu Pei-yong, Dong Jie, Xie Luo-feng,et al. Surface defect detection algorithm of magnetic tiles based on multi⁃branch convolutional neural network[J]. Journal of Jilin University (Engineering and Technology Edition),2023,53(5): 1449-1457.
3 贺秋伟,王龙山,于忠党,等. 基于图像处理和支持向量机的微型齿轮缺陷检测[J]. 吉林大学学报: 工学版, 2008, 38(3): 565-569.
He Qiu-wei, Wang Long-shan, Yu Zhong-dang,et al. Defect detection for microgear based on image processing and support vector machine[J]. Journal of Jilin University (Engineering and Technology Edition),2008, 38(3): 565-569.
4 吴凤和,崔健新,张宁,等. 基于改进YOLOv4算法的轮毂表面缺陷检测[J]. 计量学报,2022,43(11): 1404-1411.
Wu Feng-he, Cui Jian-xin, Zhang Ning,et al. Surface defect detection of wheel hub based on improved YOLOv4 algorithm[J]. Acta Metrologica Sinica,2022,43(11): 1404-1411.
5 张飞,单忠德,任永新,等. 缸盖缺陷检测系统线阵相机现场标定方法[J]. 吉林大学学报: 工学版,2017,47(2): 540-545.
Zhang Fei, Shan Zhong-de, Ren Yong-xin,et al. Calibration of line array camera for head defect detection system on cylinder cover[J]. Journal of Jilin University (Engineering and Technology Edition),2017,47(2): 540-545.
6 Zheng J F, Wu H, Han Z,et al. Insulator-defect detection algorithm based on improved YOLOv7[J]. Sensors,2022,22(22): 22228801.
7 Sohail A, Adarsh T, Medhat M,et al. A robotics inspection system for detecting defects on semi-specular painted automotive surfaces[C]∥ IEEE International Conference on Robotics and Automation,Paris,France,2020: 8928-8934.
8 Jing Z, Masaki K, Chao Z,et al. Defect inspection of coated automobile roofs using a single camera[J]. IEEJ Transactions on Electrical and Electronic Engineering,2020,15(4): 616-625.
9 Jaime M, Solanes J E, Laura A,et al. On the detection of defects on specular car body surfaces[J]. Robotics and Computer-Integrated Manufacturing,2017,48: 263-278.
10 Edris M Z B, Jawad M S, Zakaria Z. Surface defect detection and neural network recognition of automotive body panels[C]∥ International Conference on Control System,Computing and Engineering,Penang,Malaysia,2015: 117-122.
11 Yang Y X, Sun Q, Zhang D K,et al. Improved method based on faster R-CNN network optimization for small target surface defects detection of aluminum profile[C]∥ IEEE 15th International Conference on Electronic Measurement & Instruments(ICEMI), Nanjing, China, 2021: 465-470.
12 Liu M W, Lin Y H, Lo Y C,et al. Defect detection of grinded and polished workpieces using faster R-CNN[C]∥ Advanced Intelligent Mechatronics,Delft,Netherlands,2021: 1290-1296.
13 Zhang J D, Xu J B, Zhang L Y,et al. An improved mobilenet-ssd algorithm for automatic defect detection on vehicle body paint[J]. Multimedia Tools and Applications,2020,79(31/32): 23367-23385.
14 Chang F, Liu M, Dong M Y,et al. A mobile vision inspection system for tiny defect detection of smooth car-body surface based on deep ensemble learning[J]. Measurement Science and Technology,2019,30(12): 125905.
15 Chang F, Dong M Y, Liu M,et al. A lightweight appearance quality assessment system based on parallel deep learning for painted car-body[J]. IEEE Transactions on Instrumentation and Measurement,2020,69(8): 5298-5307.
16 Xu G, Guo R F, Wang H L. Quality inspection of workpiece camouflage spraying based on improved YOLOv3-tiny[C]∥ IEEE International Conference Computer and Communications,Chengdu,China,2020: 1363-1367.
17 Venkat A A, Chih H C, Hsu C C,et al. Applying deep learning to defect detection in printed circuit boards via a newest model of you-only-look-once[J]. Mathematical Biosciences and Engineering,2021,18(4): 4411-4428.
18 Liao D H, Cui Z H, Zhang X,et al. Surface defect detection and classification of Si3N4 turbine blades based on convolutional neural network and YOLOv5[J]. Advances in Mechanical Engineering,2022,14(2): 16878132221081580.
19 Tu X G, Yuan Z H, Liu B K,et al. An improved YOLOv5 for object detection in visible and thermal infrared images based on contrastive learning[J]. Frontiers in Physics,2023,11: 354-363.
20 Zhang R, Xie C, Deng L W. A fine-grained object detection model for aerial images based on YOLOv5 deep neural network[J]. Chinese Journal of Electronics,2023,32(1): 51-63.
21 Wang C Y, Alexey B, Liao H Y. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]∥ Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,Vancouver,Canada,2022: 7464-7475.
[1] 黄玲,崔躜,游峰,洪佩鑫,钟浩川,曾译萱. 适用于多车交互场景的车辆轨迹预测模型[J]. 吉林大学学报(工学版), 2024, 54(5): 1188-1195.
[2] 郭洪艳,王连冰,赵旭,戴启坤. 考虑侧向运动的整车质量与道路坡度估计[J]. 吉林大学学报(工学版), 2024, 54(5): 1175-1187.
[3] 汪少华,张启睿,施德华,殷春芳,李春. 双行星排式混合动力传动系统非线性振动响应特性分析[J]. 吉林大学学报(工学版), 2024, 54(4): 890-901.
[4] 高镇海,蔡荣贵,孙天骏,于桐,赵浩源,班浩. 人机共驾下的驾驶行为数据滤波方法[J]. 吉林大学学报(工学版), 2024, 54(3): 589-599.
[5] 谢宪毅,王禹涵,金立生,赵鑫,郭柏苍,廖亚萍,周彬,李克强. 基于改变控制时域时间步长的智能车轨迹跟踪控制[J]. 吉林大学学报(工学版), 2024, 54(3): 620-630.
[6] 邓小林,杨馥模,覃善甘. 新型仿竹六边形梯度层级多胞管耐撞性对比分析[J]. 吉林大学学报(工学版), 2024, 54(2): 333-345.
[7] 王毅刚,王玉鹏,张昊,赵思安. 高速列车转向架区域气动噪声源识别与分析[J]. 吉林大学学报(工学版), 2024, 54(2): 346-355.
[8] 聂建军,侯军凯,解晓琳,鄢鸿桢. 新型巡检机器人移动底盘设计及越障性能分析[J]. 吉林大学学报(工学版), 2024, 54(2): 356-364.
[9] 杨国俊,齐亚辉,石秀名. 基于数字图像技术的桥梁裂缝检测综述[J]. 吉林大学学报(工学版), 2024, 54(2): 313-332.
[10] 胡宏宇,张慧珺,姚荣涵,陈国迎,高菲. L3级自动驾驶接管过程驾驶员情景意识研究[J]. 吉林大学学报(工学版), 2024, 54(2): 410-418.
[11] 李雄飞,宋紫萱,朱芮,张小利. 基于多尺度融合的遥感图像变化检测模型[J]. 吉林大学学报(工学版), 2024, 54(2): 516-523.
[12] 吴骁,史文库,郭年程,赵燕燕,陈志勇,李鑫鹏,孙卓,刘健. 基于Ease off的准双曲面齿轮多目标优化[J]. 吉林大学学报(工学版), 2024, 54(1): 76-85.
[13] 陈岳林,高铸成,蔡晓东. 基于BERT与密集复合网络的长文本语义匹配模型[J]. 吉林大学学报(工学版), 2024, 54(1): 232-239.
[14] 王铁,李旭东,田程,赵宏伟. 基于多轴载荷投影构建轮辋双轴疲劳损伤模型[J]. 吉林大学学报(工学版), 2024, 54(1): 99-104.
[15] 李旭东,王新宇,田程,张新峰,牛治慧,赵志强. 基于用户关联的车辆耐久性载荷谱编制[J]. 吉林大学学报(工学版), 2024, 54(1): 66-75.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 《吉林大学学报(工学版)》编辑部
地址：长春市人民大街5988号　电话：+86-431-85095297　传真：+86-431-85094074　E-mail: xbgxb@jlu.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发