[1]王改华,李涛,吕朦,等.采用无监督学习算法与卷积的图像分类模型[J].华侨大学学报(自然科学版),2018,39(1):146-151.[doi:10.11830/ISSN.1000-5013.201703109]
　WANG Gaihua,LI Tao,Lü Meng,et al.Image Classification Model Using Unsupervised Learning Algorithm and Convolution[J].Journal of Huaqiao University(Natural Science),2018,39(1):146-151.[doi:10.11830/ISSN.1000-5013.201703109]
点击复制

采用无监督学习算法与卷积的图像分类模型()
分享到：

《华侨大学学报（自然科学版）》[ISSN:1000-5013/CN:35-1079/N]

卷:
第39卷
期数:
2018年第1期
页码:
146-151
栏目:
出版日期:
2018-01-17

文章信息/Info

Title:
Image Classification Model Using Unsupervised Learning Algorithm and Convolution
文章编号:
1000-5013(2018)01-0146-06
作者:
王改华1; 2;  李涛2;  吕朦2;  袁国亮2
1. 湖北工业大学 太阳能高效利用湖北省协同创新中心, 湖北 武汉 430068;2. 湖北工业大学 电气与电子工程学院, 湖北 武汉 430068
Author(s):
WANG Gaihua1; 2;  LI Tao2;  Lü Meng2;  YUAN Guoliang2
1. Hubei Collaborative Innovation Centre for High-Efficiency Utilization of Solar Energy, Hubei University of Technology, Wuhan 430068, China; 2. School of Electrical and Electronic Engineering, Hubei University of Technology, Wuhan 430068, China
关键词:
K-means聚类;  图像分类;  卷积;  卷积神经网络;  Dropout
Keywords:
K-means clustering;  image classification;  convolution;  convolutional neural net;  Dropout
分类号:
TP391.41
DOI:
10.11830/ISSN.1000-5013.201703109
文献标志码:
A
摘要:
为了提高图像分类精度,降低训练复杂度,提出一种采用无监督学习算法与卷积构造的图像分类模型.首先,从输入无标签图像中随机抽取大小相同的图像块构成数据集,进行预处理.其次,将预处理后的图像块通过两次K-means聚类算法提取字典,并采用离散卷积操作提取最终图像特征.最后,采用Softmax分类器对提取的图像特征进行分类,得出准确率.将该模型与卷积神经网络(CNN),Dropout CNN网络进行比较,结果表明:在对大规模高维图像分类上,文中模型具有分类精确度高、简单、训练参数少、适应度高等优点.
Abstract:
To improve image classification accuracy and reduce the training complexity, a image classification model based on unsupervised learning algorithm and convolution is proposed. Above all, the dataset is composed of randomly extracted image patch of the same size from the input unlabeled images, and the dataset was pre-processing. Next, the preprocessed image patch is fed into twice K-means clustering algorithm to extract the dictionary, and the final image feature is extracted by the discrete convolution operation. Finally, the accuracy of image classification was obtained by classifying the extracted image features using Softmax classifier. Compared with the convolutional neural net(CNN)and Dropout CNN. The results shows that the proposed model has the advantages of high classification accuracy, simplicity, less training parameters and high adaptability for large-scale high-dimensional image classification.

参考文献/References:

[1] CIRESAN D,MEIER U,SCHMIDHUBER J.Multi-column deep neural networks for image classificatiion[C]//IEEE Conference on Computer Vision and Pattern Recognition.Rhode lsland:IEEE Press,2012:3642-3649.
[2] SHI Baoguang,BAI Xiang,YAO Cong.Script identification in the wild via discriminative convolutional neural network[J].Pattern Recognition,2016,52:448-458.DOI:10.1016/j.patcog.2015.11.005.
[3] CHATFIELD K,SIMONYAN K,VEDALDI A,et al.Return of the devil in the details: Delving deep into convolutional nets[C]//Proc of the British Machine Vision Conference.Nottingham:BMVA Press,2014:1-12.DOI:10.5244/C.28.6.
[4] SIMARD P Y,STEINKRAUS D,PLATT J C.Best practices for convolutional neural networks applied to visual document analysis[C]//Proc of the 17th International Conference on Document Analysis and Recognition.Edinburgh:IEEE Press,2013:958-963.
[5] SUN Yi,WANG Xiaogang,TANG Xiaoou.Deep convolutional network cascade for facial point detection[C]//Proc of the 26th IEEE Conference on Computer Vision and Pattern Recognition.Oregon:IEEE Press,2013:3476-3483.
[6] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.Imagenet classification with deep convolutional neural networks[C]//Proc of the 19th International Conference on Nerual Information Processing.New York:ACM Press,2012:1097-1105.DOI:10.1145/3065386.
[7] ZEILER M D,TAYLOR G W,FERGUS R.Adaptivedeconvolutional networks for mid and high level feature learning[C]//Proc of the International Conference on Computer Vision.Barcelona:IEEE Press,2011:2018-2025.
[8] ZEILER M D,FERGUS R.Visualizing and understanding convolutional networks[C]//Proc of the European Conference on Computer Vision.Berlin:Springer,2014:818-833.
[9] SERMANET P,EIGEN D,ZHANG Xiang,et al.Overfeat: Integrated recognition, localization and detection using convolutional networks[J].Eprint Arxiv,2014(V4):1-16.
[10] JIE Zequn,YAN Shuicheng.Robust scene classification with cross-level LLC coding on CNN features[M].Bangalore:Springer,2014.
[11] BARAT C,DUCOTTET C.String representations and distances in deep convolutional neural networks for image classification[J].Pattern Recognition,2016,54(C):104-115.DOI:10.1016/j.patcog.2016.01.007.
[12] 吕娜.基于深度层次特征学习的大规模图像分类研究[D].成都:电子科技大学,2015.
[13] COATES A,NG A Y,LEE H.An analysis of single-layer networks in unsupervised feature learning[J].Journal of Machine Learning Research,2011,15:215-233.
[14] 何鹏程.改进的卷积神经网络模型及其应用研究[D].大连:大连理工大学,2015.
[15] PALM R B.Prediction as a candidate for learning deep hierarchical models of data[D].Denmark:Technical University of Denmark,2012.
[16] HINTON G E,SRIVASTAVA N,KRIZHEVSKY A,et al.Improving neural networks by preventing co-adaptation of feature detectors[J].Computer Science,2012,3(4):213-223.

备注/Memo

备注/Memo:
收稿日期: 2017-03-04
通信作者: 王改华(1979-),女,讲师,博士,主要从事模式识别与图像处理的研究.E-mail:wanggh@aliyun.com.
基金项目: 湖北省教育厅科学技术研究青年项目(Q20161405); 湖北省教育厅科学技术研究指导性项目(B2015045)
更新日期/Last Update: 2018-01-20