| 摘 要: 针对现有的PCB板缺陷检测方法存在参数量大、精确度较低等问题,提出了一种改进的 YOLOv8缺陷检测与识别方法。首先,对BiFPN双向特征金字塔网络进行改进,在BiFPN的基础上整合P2特征层强化对小型缺陷目标的检测能力,并优化网络连接结构以适配 YOLOv8主干网络;并添加接近无参注意力 TA(TokenAttention)提高模型识别准确度。其次,设计全新的CSC(Channel-SpatialCompression)模块代替Backbone的C2f模块,减少模型参数冗余。最后,将模型均值平均精度提升至98.9%。在PKU-Market-PCB数据集上的实验结果表明,与原算法相比,改进算法的mAP 提升了3.2%,在PCB缺陷检测任务中展现出显著优势。 |
| 关键词: 机器视觉 YOLOv8算法 小目标检测 多尺度 缺陷检测 |
|
中图分类号: TP391.4
文献标识码: A
|
|
| PCB Defect Detection Method Based on BST-YOLOv8 |
|
ZUO Guanfang1,2, HUANG Tao1, SHA Yanyou1, XIA Yuchen1
|
(1.School of Electronics and Information Engineering, Nanjing University of Information Science & Technology, Nanjing 210044, China;
2.School of Electronic Information Engineering, Wuxi University, Wuxi 214105, China)
zgf@cwxu.edu.cn; 597912118@qq.com; 1819735834@qq.com; 3239248402@qq.com
|
| Abstract: To address the issues of large parameter size and low accuracy in existing PCB defect detection methods, this paper proposes an improved YOLOv8-based defect detection and recognition approach. The Bidirectional Feature Pyramid Network (BiFPN) is enhanced by integrating the P2 feature layer to improve the detection accuracy of small objects, and the network connections are modified to better suit the YOLOv8 architecture. Additionally, a nearly paramete-r free Token Attention (TA) mechanism is incorporated to boost model recognition accuracy. Furthermore, a novel CSC module is designed to replace the C2f module in the Backbone, reducing parameter redundancy and enabling faster and more precise target localization. Evaluations on the PKU-Marke-t PCB dataset show that the
improved algorithm achieves a 3.2% increase in mean Average Precision (mAP) compared to the original method,demonstrating significant effectiveness in PCB defect detection. |
| Keywords: machine vision YOLOv8 algorithm small object detection mult-i scale defect detection |
