期刊
IEEE TRANSACTIONS ON COMPONENTS PACKAGING AND MANUFACTURING TECHNOLOGY
卷 12, 期 2, 页码 217-227出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TCPMT.2021.3136823
关键词
Feature extraction; Convolution; Training; Semantics; Object detection; Manufacturing; Convolutional neural networks; Dataset enhancement; faster R-CNN; feature pyramid networks; focal hard samples
类别
资金
- National Natural Science Foundation of China [62176034, 61905033]
- National Natural Science Foundation of Chongqing [cstc2021jcyj-msxmX0941]
- Scientific and Technological Research Program of Chongqing Municipal Education Commission [KJQN202101907]
- Doctoral Talent Training Project of Chongqing University of Posts and Telecommunications [BYJS202007]
This article presents a PCB defect detection algorithm based on the extended feature pyramid network model. By incorporating multiscale fusion and introducing focal loss function, the algorithm addresses the challenge of small object detection. Experimental results show that the algorithm achieves a mean average precision (mAP) of 96.2% on the public PCB dataset, surpassing the state-of-the-art methods.
Suffering from the diversity, complexity, and miniaturization of printed circuit board (PCB) defects, traditional detection methods are difficult to detect. Despite object detection has made significant advances based on deep neural networks, it remains a challenge to focus on small objects. We address this challenge by allowing multiscale fusion. We introduce a PCB defect detection algorithm based on extended feature pyramid network model in this article. The backbone is constructed by part of ResNet-101, in order to accurately locate and identify small objects, this article constructs a feature layer, which integrates high-level semantic information and low-level geometric information. Based on feature pyramid networks (FPN) network structure, using 1x1convolution lateral fusion of the previous semantic information, the fused features use 3x3 convolution to obtain the final feature layer. The problem that PCB defects are difficult to classify is considered, the focal loss function is introduced. To reduce over-fitting in the training process, the original data are enhanced using image clipping and rotation. Through the quantitative analysis on PCB defect dataset, these results are the best to be used in fused low-level feature layer for detection of the mean average precision (mAP). This is 96.2% on the public PCB dataset, which is surpassing the state-of-the-art methods.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据