基于改进YOLOv7的PCB缺陷检测算法

doi:10.19678/j.issn.1000-3428.0068588

摘要/Abstract

摘要：

在PCB缺陷检测领域中检测精度的提高一直是1个具有挑战性的任务。为了解决这个问题, 提出一系列基于PCB缺陷检测的改进方法。首先, 引入一种新的注意力机制, 即BiFormer注意力机制, 这种机制利用双层路由实现动态的稀疏注意力, 从而减少计算量; 其次, 采用一种创新的上采样算子CARAFE, 能够结合语义信息与内容信息进行上采样, 使得上采样过程更加全面且高效; 最后, 基于MPDIoU度量采用一种新的损失函数, 即LMPDIoU损失函数, 能够有效地处理不平衡类别、小目标和密集性问题, 从而进一步提高图像检测的性能。实验结果表明, 所提改进后的模型在平均精度均值(mAP)方面取得了显著提高, 达到了93.91%, 与原YOLOv5模型相比提高了13.12个百分点, 同时, 在识别精度方面, 所提改进后的模型表现也非常出色, 达到了90.55%, 与原YOLOv5模型相比提高了8.74个百分点。引入BiFormer注意力机制、CARAFE上采样算子以及LMPDIoU损失函数, 对于提高PCB缺陷检测的精度和效率具有非常积极的作用, 为工业检测领域的研究提供了有价值的参考。

关键词: PCB缺陷, BiFormer注意力机制, MPDIoU损失函数, 上采样算子CARAFE, 目标检测

Abstract:

Achieving enhanced detection accuracy is a challenging task in the field of PCB defect detection. To address this problem, this study proposes a series of improvement methods based on PCB defect detection. First, a novel attention mechanism, referred to as BiFormer, is introduced. This mechanism uses dual-layer routing to achieve dynamic sparse attention, thereby reducing the amount of computation required. Second, an innovative upsampling operator called CARAFE is employed. This operator combines semantic and content information for upsampling, thereby making the upsampling process more comprehensive and efficient. Finally, a new loss function based on the MPDIoU metric, referred to as the LMPDIoU loss function, is adopted. This loss function effectively addresses unbalanced categories, small targets, and denseness problems, thereby further improving image detection performance. The experimental results reveal that the model achieves a significant improvement in mean Average Precision (mAP) with a score of 93.91%, 13.12 percentage points higher than that of the original model. In terms of recognition accuracy, the new model reached a score of 90.55%, representing an improvement of 8.74 percentage points. These results show that the introduction of the BiFormer attention mechanism, CARAFE upsampling operator, and LMPDIoU loss function effectively improves the accuracy and efficiency of PCB defect detection. Thus, the proposed methods provide valuable references for research in industrial inspection, laying the foundation for future research and applications.

Key words: PCB defect, BiFormer attention mechanism, MPDIoU loss function, upsampling operator CARAFE, target detection

张旭, 陈慈发, 董方敏. 基于改进YOLOv7的PCB缺陷检测算法[J]. 计算机工程, 2024, 50(12): 318-328.

ZHANG Xu, CHEN Cifa, DONG Fangmin. PCB Defect Detection Algorithm Based on Improved YOLOv7[J]. Computer Engineering, 2024, 50(12): 318-328.

https://www.ecice06.com/CN/Y2024/V50/I12/318

图/表 16

图1 YOLOv7结构

Fig.1 YOLOv7 structure

图2 BRA结构

Fig.2 BRA structure

图3 BiFormer的整体架构

Fig.3 Overall architecture of BiFormer

图4 CARAFE整体框架

Fig.4 Overall architecture of CARAFE

图5 IoU和GIoU公式计算示意图

Fig.5 Schematic diagram of IoU and GIoU formula calculation

图6 DIoU和CIoU公式计算示意图

Fig.6 Schematic diagram of DIoU and CIoU formula calculation

图7 长宽比相同示例

Fig.7 Example of same aspect ratio

图8 改进后各缺陷的混淆矩阵

Fig.8 Confusion matrix of each defect after improvement

图9 原YOLOv7和改进YOLOv7模型的简单目标检测效果

Fig.9 Simple object detection effect of original YOLOv7 and improved YOLOv7 models

图10 原YOLOv7和改进YOLOv7模型的密集目标检测效果

Fig.10 The dense object detection effect of the original YOLOv7 and improved YOLOv7 models

图11 原YOLOv7和改进YOLOv7模型综合性能对比

Fig.11 Comparison of comprehensive performance between original YOLOv7 and improved YOLOv7 models

图12 泛化数据示例展示

Fig.12 Example display of generalized data

参考文献 25

1	吴一全, 赵朗月, 苑玉彬, 等. 基于机器视觉的PCB缺陷检测算法研究现状及展望. 仪器仪表学报, 2022, 43 (8): 1- 17. doi: 10.19650/j.cnki.cjsi.J2209701
	WU Y Q , ZHAO L Y , WAN Y B , et al. Research status and the prospect of PCB defect detection algorithm based on machine vision. Chinese Journal of Scientific Instrument, 2022, 43 (8): 1- 17. doi: 10.19650/j.cnki.cjsi.J2209701
2	陈新, 陈云, 杨志军, 等. 电子器件高精高效制造的若干关键技术研究进展. 机械工程学报, 2023, 59 (19): 164- 175. doi: 10.3901/JME.2023.19.164
	CHEN X , CHEN Y , YANG Z J , et al. Research advances in several key technologies for high precision and high efficiency manufacturing of electronic devices. Journal of Mechanical Engineering, 2023, 59 (19): 164- 175. doi: 10.3901/JME.2023.19.164
3	阮光册, 钟静涵, 张祎笛. 基于深度学习的术语识别研究综述. 数据分析与知识发现, 2024, 8 (4): 64- 75. doi: 10.11925/infotech.2096-3467.2023.0158
	RUAN G S , ZHONG J H , ZHANG W D . Review of term recognition studies based on deep learning. Data Analysis and Knowledge Discovery, 2024, 8 (4): 64- 75. doi: 10.11925/infotech.2096-3467.2023.0158
4	李正明, 黎宏, 孙俊. 基于数字图像处理的印刷电路板缺陷检测. 仪表技术与传感器, 2012, (8): 87- 89. doi: 10.3969/j.issn.1002-1841.2012.08.030
	LI Z M , LI H , SUN J . Detection of PCB based on digital image processing. Instrument Technique and Sensor, 2012, (8): 87- 89. doi: 10.3969/j.issn.1002-1841.2012.08.030
5	罗嗣卿, 陈慧. 基于生成对抗网络的图像场景转换. 计算机工程, 2023, 49 (4): 217- 225. doi: 10.19678/j.issn.1000-3428.0066077
	LUO S Q , CHEN H . Image-scene transformation based on generative adversarial networks. Computer Engineering, 2023, 49 (4): 217- 225. doi: 10.19678/j.issn.1000-3428.0066077
6	李刚, 邵瑞, 周鸣乐, 等. 基于注意力的轻量级工业产品缺陷检测网络. 计算机工程, 2023, 49 (11): 275- 283. doi: 10.19678/j.issn.1000-3428.0066270
	LI G , SHAO R , ZHOU M L , et al. Lightweight industrial products defect detection network based on attention. Computer Engineering, 2023, 49 (11): 275- 283. doi: 10.19678/j.issn.1000-3428.0066270
7	熊光洁, 马树元, 聂学俊, 等. 基于机器视觉的高密度电路板缺陷检测系统. 计算机测量与控制, 2011, 19 (8): 1824- 1826. URL
	XIONG G J , MA S Y , NIE X J , et al. Defects inspection system of HID PCB based on machine vision. Computer Measurement & Control, 2011, 19 (8): 1824- 1826. URL
8	于洋, 肖扬, 任芳语, 等. 基于OpenCV与遥感特征图像提取与配准方法综述. 科技创新导报, 2018, 15 (10): 48- 49. doi: 10.16660/j.cnki.1674-098X.2018.10.048
	YU Y , XIAO Y , REN F Y , et al. Overview of OpenCV and remote sensing feature image extraction and registration methods. Science and Technology Innovation Herald, 2018, 15 (10): 48- 49. doi: 10.16660/j.cnki.1674-098X.2018.10.048
9	文韬, 周稻祥, 李明. Mask R-CNN中特征不平衡问题的全局信息融合方法. 计算机工程, 2021, 47 (3): 256-260, 268. doi: 10.19678/j.issn.1000-3428.0056469
	WEN T , ZHOU D X , LI M . Global information fusion method for feature imbalance problem in Mask R-CNN. Computer Engineering, 2021, 47 (3): 256-260, 268. doi: 10.19678/j.issn.1000-3428.0056469
10	康劲松, 周艳萍, 孙梁榕, 等. 激光无线电能传输系统对准环节设计. 光电工程, 2023, 50 (7): 230109. doi: 10.12086/oee.2023.230109
	KANG J S , ZHOU Y P , SUN L R , et al. Design of alignment subsystem for laser wireless power transmission system. Opto-Electronic Engineering, 2023, 50 (7): 230109. doi: 10.12086/oee.2023.230109
11	许思昂, 李艺杰, 梁桥康, 等. 基于改进YOLOv5算法的PCB裸板缺陷检测. 包装工程, 2022, 43 (15): 33- 41. doi: 10.19554/j.cnki.1001-3563.2022.15.004
	XU S A , LI Y J , LIANG Q K , et al. Bare PCB defect detection based on improved YOLOv5 algorithm. Packaging Engineering, 2022, 43 (15): 33- 41. doi: 10.19554/j.cnki.1001-3563.2022.15.004
12	邓天民, 谭思奇, 蒲龙忠. 基于改进YOLOv5s的交通信号灯识别方法. 计算机工程, 2022, 48 (9): 55- 62. doi: 10.19678/j.issn.1000-3428.0062843
	DENG T M , TAN S Q , PU L Z . Traffic light recognition method based on improved YOLOv5s. Computer Engineering, 2022, 48 (9): 55- 62. doi: 10.19678/j.issn.1000-3428.0062843
13	阴晓阳. 改进YOLOv3的PCB表面缺陷检测方法研究[D]. 沈阳: 辽宁工程技术大学, 2023. URL
	YIN X Y. Research on PCB surface defect detection method based on improved YOLOv3[D]. Shenyang: Liaoning Technical University, 2023. (in Chinese)
14	张先炼, 王国杰. 混合遗传算法综述. 电子世界, 2015, (14): 120- 121. doi: 10.3969/j.issn.1003-0522.2015.14.070
	ZHANG X L , WAGN G J . Overview of hybrid genetic algorithm. Electronics World, 2015, (14): 120- 121. doi: 10.3969/j.issn.1003-0522.2015.14.070
15	芦照烜, 朱晓龙, 龙顺宇, 等. 基于Cascade R-CNN改进的PCB瑕疵检测方法. 现代电子技术, 2023, 46 (15): 172- 179. doi: 10.16652/j.issn.1004-373x.2023.15.030
	LU Z X , ZHU X L , LONG S Y , et al. Improved PCB defect detection method based on Cascade R-CNN. Modern Electronics Technique, 2023, 46 (15): 172- 179. doi: 10.16652/j.issn.1004-373x.2023.15.030
16	ZHENG Q H , ZHAO P H , LI Y , et al. Spectrum interference-based two-level data augmentation method in deep learning for automatic modulation classification. Neural Computing and Applications, 2021, 33 (13): 7723- 7745. doi: 10.1007/s00521-020-05514-1
17	WANG C Y, BOCHKOVSKIY A, LIAO H Y M. 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. Washington D. C., USA: IEEE Press, 2023: 7464-7475. URL
18	朱贺, 杨华, 尹周平. 纹理表面缺陷机器视觉检测方法综述. 机械科学与技术, 2023, 42 (8): 1293- 1315. doi: 10.13433/j.cnki.1003-8728.20220086
	ZHU H , YANG H , YIN Z P . Review of machine vision detection methods for texture surface defects. Mechanical Science and Technology for Aerospace Engineering, 2023, 42 (8): 1293- 1315. doi: 10.13433/j.cnki.1003-8728.20220086
19	ZHU L, WANG X, KE Z, et al. BiFormer: vision transformer with bi-level routing attention[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2023: 10323-10333. URL
20	祁宣豪, 智敏. 图像处理中注意力机制综述. 计算机科学与探索, 2024, (2): 1- 20. doi: 10.3778/j.issn.1673-9418.2305057
	QI X H , ZHI M . Review of attention mechanisms in image processing. Journal of Frontiers of Computer Science and Technology, 2024, (2): 1- 20. doi: 10.3778/j.issn.1673-9418.2305057
21	朱红红, 刘雅, 杨希望. 传统优化方法与机器学习融合的研究进展[C]//第十八届中国管理学年会暨"一带一路"十周年研讨会. 乌鲁木齐: 中国管理现代化研究会, 2023: 1-10.
	ZHU H H, LIU Y, YANG X W. Research reviews of the integration of traditional optimization methods and machine learning[C]//Proceedings of the 18th China Management Annual Conference and the 10th Anniversary Seminar of the "the Belt and Road". Urumqi: Chinese Academy of Management Annual Meeting, 2023: 1-10. (in Chinese)
22	魏陈浩, 杨睿, 刘振丙, 等. 具有双层路由注意力的YOLOv8道路场景目标检测方法. 图学学报, 2023, 44 (6): 1104- 1111. doi: 10.11996/JG.j.2095-302X.2023061104
	WEI C H , YANG R , LIU Z B , et al. YOLOv8 with bi-level routing attention for road scene object detection. Journal of Graphics, 2023, 44 (6): 1104- 1111. doi: 10.11996/JG.j.2095-302X.2023061104
23	WANG J Q, CHEN K, XU R, et al. CARAFE: content-aware reassembly of features[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2019: 3007-3016. URL
24	MA S L, XU Y. MPDIoU: a loss for efficient and accurate bounding box regression[EB/OL].[2023-09-12]. https://arxiv.org/abs/2307.07662.
25	张翠文, 张长伦, 何强, 等. 目标检测中框回归损失函数的研究. 计算机工程与应用, 2021, 57 (20): 97- 103. doi: 10.3778/j.issn.1002-8331.2009-0450
	ZHANG C W , ZHAGN C L , HE Q , et al. Research on loss function of box regression in object detection. Computer Engineering and Applications, 2021, 57 (20): 97- 103. doi: 10.3778/j.issn.1002-8331.2009-0450

[1]	张天鹏, 韩晶, 吕学强. 基于多任务学习的超分辨率辅助小目标检测[J]. 计算机工程, 2024, 50(9): 304-312.
[2]	曾钰琦, 刘博, 钟柏昌, 钟瑾. 智慧教育下基于改进YOLOv8的学生课堂行为检测算法[J]. 计算机工程, 2024, 50(9): 344-355.
[3]	王昱婷, 刘志明, 万亚平, 朱涛. 基于可见光与红外图像的弱光条件下目标检测[J]. 计算机工程, 2024, 50(8): 270-281.
[4]	贵向泉, 刘世清, 李立, 秦庆松, 李唐艳. 基于改进YOLOv8的景区行人检测算法[J]. 计算机工程, 2024, 50(7): 342-351.
[5]	刘仕兵, 周诗涵. 高铁接触网绝缘子检测算法研究[J]. 计算机工程, 2024, 50(5): 200-208.
[6]	孙帮勇, 马铭, 于涛. 基于区域特征强化的多尺度伪装目标检测方法[J]. 计算机工程, 2024, 50(5): 209-219.
[7]	赵继达, 甄国涌, 储成群. 基于YOLOv8的无人机图像目标检测算法[J]. 计算机工程, 2024, 50(4): 113-120.
[8]	崔丽群, 曹华维. 基于改进YOLOv5的遥感图像目标检测[J]. 计算机工程, 2024, 50(4): 228-236.
[9]	周金涛, 高迪驹, 刘志全. 基于全景视觉的无人船水面障碍物检测方法[J]. 计算机工程, 2024, 50(2): 113-121.
[10]	任书玉, 汪晓丁, 林晖. 目标检测中注意力机制综述[J]. 计算机工程, 2024, 50(12): 16-32.
[11]	李沼洁, 朱恒亮, 毛国君, 杨鑫. 渐进式特征增强的弱监督显著性目标检测[J]. 计算机工程, 2024, 50(12): 233-244.
[12]	王林, 赵莉, 王无为. 高动态场景下无人机空对空目标检测[J]. 计算机工程, 2024, 50(12): 265-275.
[13]	徐守坤, 张路军, 石林, 刘毅. 意图注意力引导的小样本3D点云目标检测[J]. 计算机工程, 2024, 50(12): 288-295.
[14]	谭若琦, 董明刚, 赵唯肖, 武天昊. 基于语义对齐和层次优化的非机动车车牌识别定位方法[J]. 计算机工程, 2024, 50(11): 142-151.
[15]	刘宏纬, 邵东恒, 杨剑, 魏宪, 李科, 游雄. 基于鸟瞰图融合的多级旋转等变目标检测网络[J]. 计算机工程, 2024, 50(11): 246-257.

选择文件类型/文献管理软件名称

选择包含的内容