基于卷积神经网络的电路板元器件分类算法

详细信息    查看全文 | 推荐本文 |

英文篇名：Circuit board component classification algorithm based on convolutional neural network 作者：李威 ; 张文亮 ; 蔡立明 ; 余正海 ; 段荣 英文作者：LI Wei;ZHANG Wen-Liang;CAI Li-Ming;YU Zheng-Hai;DUAN Rong;Beijing Institute of Aerospace Control Devices; 关键词：电路板板载元器件 ; 监督学习算法 ; 直方图均衡化 ; 卷积神经网络 英文关键词：circuit board components;;supervised learning algorithm;;histogram equalization;;convolution neural network 中文刊名：黑龙江大学工程学报 英文刊名：Journal of Engineering of Heilongjiang University 机构：北京航天控制仪器研究所; 出版日期：2019-09-25 出版单位：黑龙江大学工程学报 年：2019 期：03 基金：国家自然科学基金资助项目(61571145) 语种：中文; 页：80-83 页数：4 CN：23-1566/T ISSN：2095-008X 分类号：TP183;TN41;TP391.41

摘要

为了实现电路板板载元器件的非接触及快速准确识别,提出了一种基于卷积神经网络的电路板元器件分类算法(C-CNN)。将工业相机拍摄的电路板图像进行直方图均衡化,减少光照不均产生的影响。提取监督样本用于训练卷积神经网络,并创建分类模型。将待测图像输入深度卷积神经网络进行特征提取,并利用分类模型对电路板元器件进行识别。利用实际拍摄的电路板板载元器件图像数据集验证文章提出的算法,实验结果表明:C-CNN能够对元器件进行识别,并且性能优于支持向量机及深度神经网络算法。
        In order to implement non-contact and fast and accurate identification of circuit board components, a classification algorithm of circuit board components based on convolution neural network(C-CNN) is proposed in this paper. Firstly, histogram equalization of circuit board images taken by industrial cameras is used to reduce the impact of uneven illumination. Supervisory samples are extracted to train convolutional neural networks, and a classification model is created. Then, the image under detection is input into the depth convolution neural network for feature extraction, and the classification model is used to identify the circuit board components. The experimental results show that C-CNN can recognize components, and its performance is better than that of support vector machine and deep neural network.

引文

[1] 张顺辽,苏明,钟立,等.基于图像处理的印制电路板缺陷自动检测[J].计算机测量与控制,2007,15(5):591-592.
    [2] 闫梦涛,苏玮,冉海周.基于机器视觉的 PCBA 元器件实时检测系统[J].无线电工程,2018,48(4):272-278.
    [3] 张银玲,沈永良,张梦营.基于机器视觉的螺纹缺陷检测方法[J].黑龙江大学工程学报,2016,7(4):92-97.
    [4] 尤海云,王绍纯.形态学细化算法在印制电路板PCB定位中的应用[J].自动化技术与应用,2003,22(10):19-21.
    [5] 李汉国,何星,阎晓娜.印制电路板的自动光学检测[J].无损检测,2004,26(6):307-309.
    [6] 胡跃民,谭颖.自动光学检测在中国的应用现状和发展[J].微计算机信息,2006,22(2):143-146.
    [7] Lecun Y,Jackel L D,Bottou L,et al.Comparison of learing algorithms for handwritten digit recognition[C]//Proceedings of the International Conference on Artificial Neural Networks.Paris,France,1995,60:53-60.
    [8] 郭睿,史小平,贾殿坤.基于深度卷积网络的图像超分辨率重建[J].黑龙江大学工程学报,2018,9(4):52-59.
    [9] Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Boston,USA,2015:3431-3440.
    [10] 彭明生,莫玉龙.基于数学形态和神经网络的纹理分类[J].上海大学学报:自然科学版,1999,5(2):169-171.
    [11] 王健,李孝虔.一种基于特征组合和卷积神经网络的心脏病预测新方法[J].黑龙江大学自然科学学报,2019,36(11):115-120.