基于改进YOLOv11的PCB缺陷检测研究

Study on PCB defect detection based on improved YOLOv11

摘要

针对印刷电路板(PCB,Printed Circuit Board)表面缺陷检测中YOLOv11模型存在骨干网络计算成本高、特征提取能力不足及小目标检测准确率有限的问题,提出一种改进的YOLOv11模型。该模型通过引入C3k2_iRMB_Cascaded模块,融合级联组注意力(Cascaded Group Attention,CGA)与倒置残差移动块(Improved Residual Mobile Block,iRMB),增强多尺度特征提取能力的同时降低计算复杂度;采用选择性边界聚合(Selective Boundary Aggregation,SBA)模块构建重校准特征金字塔网络(Re-Calibration FPN),通过双向特征融合与自适应注意力机制解决传统FPN的语义丢失问题,提升小目标检测精度;设计Detect_LSCD(Lightweight Shared Convolutional Detection)轻量检测头,结合共享卷积与尺度自适应缩放机制,优化多尺度目标检测的鲁棒性。实验结果表明,改进模型较原始YOLOv11n模型P_mAP0.5提升3.4%,P_mAP0.5:0.95提升1.5%,有效平衡了检测精度与实时性,满足现代电子制造系统的质量检测需求。

Abstract

To address the issues of high computational cost in the backbone network, insufficient feature extraction capability, and limited accuracy in small target detection in the YOLOv11 model for printed circuit board (PCB) surface defect detection, an improved YOLOv11 model is proposed. The model introduces the C3k2_iRMB_Cascaded module, which integrates Cascaded Group Attention (CGA) and Inverted Residual Mobile Block (iRMB) to enhance multi-scale feature extraction while reducing computational complexity. A Selective Boundary Aggregation (SBA) module is adopted to construct a Re-Calibration Feature Pyramid Network (Re-Calibration FPN), addressing the semantic loss issue in traditional FPN through bidirectional feature fusion and an adaptive attention mechanism, thereby improving small target detection accuracy. A lightweight Detect_LSCD detection head is designed, incorporating shared convolution and a scale-adaptive scaling mechanism to enhance the robustness of multi-scale object detection. Experimental results demonstrate that the improved model achieves a 3.4% increase in P_mAP0.5 and a 1.5% increase in P_mAP0.5:0.95 compared with the original YOLOv11n model, effectively balancing detection accuracy and real-time performance, making it suitable for quality inspection requirements in modern electronic manufacturing systems.

关键词

印刷电路板 / YOLOv11 / 缺陷检测 / 小目标检测

Key words

printed circuit board / YOLOv11 / defect detection / small target detection

参考文献

[1] JIANG Y Y, CAI M N. Lightweight PCB defect detection network Multi—CRYOLO[J]. Journal of Electronic Measurement and Instrumentation, 2023, 37(11): 217-224.

[2] 郭民, 王蕊. AOI技术在PCB缺陷检测中的设计与实现[J]. 测控技术, 2016, 35(12): 127-130.

[3] ANITHA D, RAO M. A survey on defect detection in bare PCB and assembled PCB using image processing techniques[C]// International Conference on Wireless Communications, Signal Processing and NETWORKING, Chennai: IEEE, 2018: 39-43.

[4] YILIN M, ZHEWEI L, XIANGNING W, et al. Cost—sensitive siamese network for PCB defect classification[J]. Computational Intelligence and Neuroscience, 2021(1): 7550670.

[5] 伍济钢, 梁谋, 曹鸿, 等. 基于改进YOLOv5的PCB小目标缺陷检测研究[J]. 光电子·激光, 2024, 35(2): 155-163.

[6] 解琳, 韩跃平, 翟倩, 等. 基于改进YOLOv7—tiny的PCB表面缺陷检测[J]. 测试技术学报, 2025, 39(1): 81-87.

[7] 吕秀丽, 杨昕升, 曹志民. 改进YOLOv8的PCB表面缺陷检测算法[J]. 电子测量技术, 2024, 47(12): 100-108.

[8] 彭在欢, 常光超, 任传成. 基于DLDE—YOLOv10n的PCB板表面缺陷检测算法[J]. 宁夏师范大学学报, 2025, 46(4): 60-70.

[9] 王欣璐, 郑晓亮, 来文豪. 基于TAC—YOLOv11s的PCB缺陷检测与实例分割算法[J]. 湖北民族大学学报(自然科学版), 2025, 43(1): 80-85.

[10] ZHANG J, LI X, LI J, et al. Rethinking mobile block for efficient attention—based models[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision, New York: IEEE, 2023: 1389-1400.

[11] LIU X, PENG H, ZHENG N, et al. EfficientViT: Memory efficient vision transformer with cascaded group attention[J]. ArXiv, 2023, abs/2305.07027. DOI: 10.48550/arXiv.2305.07027.

[12] BI J, LI K, ZHENG X, et al. SPDC—YOLO: An efficient small target detection network based on improved YOLOv8 for drone aerial image[J]. Remote Sensing, 2025, 17(4): 685.

[13] 吴斌斌, 张礼华, 刘军伟, 等. 基于改进的EP—RTDETR小目标PCB表面缺陷检测[J]. 制造技术与机床, 2025(3): 139-148.

基金资助

安徽省高校科学研究项目(2024AH050213)

华为·安徽2025产学合作创新课题项目(JXGG-02)