您的位置:山东大学 -> 科技期刊社 -> 《山东大学学报(工学版)》

山东大学学报 (工学版) ›› 2015, Vol. 45 ›› Issue (5): 13-21.doi: 10.6040/j.issn.1672-3961.2.2015.168

• • 上一篇    

基于概率密度分布一致约束的最小最大概率机图像分类算法

王晓初1,王士同1,包芳2   

  1. 1.江南大学数字媒体学院, 江苏 无锡 214122;2.江阴职业技术学院计算机系, 江苏 江阴 214405
  • 发布日期:2020-05-26
  • 通讯作者: 王晓初(1987- ),男,山西大同人,硕士研究生,主要研究方向为模式识别,数字图像处理.E-mail:icnice@yeah.net
  • 作者简介:王士同(1964- ),男,江苏扬州人,教授,硕士,主要研究方向为人工智能,模式识别,生物信息.E-mail: wxwangst@aliyun.com. *通信作者:王晓初(1987- ),男,山西大同人,硕士研究生,主要研究方向为模式识别,数字图像处理.E-mail:icnice@yeah.net
  • 基金资助:
    国家自然科学基金资助项目(61170122,61272210)

Image classification algorithm based on minimax probability machine with regularized probability density concensus

WANG Xiaochu1, WANG Shitong1, BAO Fang2   

  1. 1. School of Digital Media, Jiangnan University, Wuxi 214122, Jiangsu, China;
    2. Department of Computer Science, Jiangyin Pdyteehnie College, Jiangyin 214405, Jiangsu, China
  • Published:2020-05-26

摘要: 为了解决含有大量未标记样本的图像分类问题,提出了基于概率密度分布一致约束的最小最大概率机图像分类算法(image classification algorithm based on minimax probability machine regularized by probability density concensus, PDMPM)。用概率密度估计函数对标记图像样本和未标记图像样本在超平面所在空间的分布进行估计,最小化标记图像样本和未标记图像样本在超平面所在空间的分布差异,得到概率密度估计约束项。把概率密度估计约束项融入到非线性最小最大概率机并用于图像分类。在耶鲁大学人脸数据库、加利福尼亚理工学院101类图像数据库的5类和15场景数据库中的10类分类准确率的试验中,该算法较非线性最小最大概率机大约平均提高了3.99%,从而表明该方法充分利用了未标记图像样本包含的分布信息来约束标记图像样本的分布,使得图像分类超平面更加准确。

关键词: 图像分类, 未标记样本, 概率密度估计, 分类超平面, 最小最大概率机

Abstract: In order to solve image classification problem of which the images contained labeled and unlabeled samples, this research proposed an image classification algorithm based on minimax probability machine regularized by probability density concensus(called PDMPM). The distribution of the image samples in the hyperplane was estimated by using the probability density estimation function and probability density estimation constrained item was got by minimizing the distribution of the labeled and unlabeled samples. The probability density estimation constraint item was integrated into the nonlinear minimax maximum probability machine and used for image classification. The accuracy of the proposed algorithm was increased by 3.99% compared with Gaussian kernel minimax probability machine in the test of Yale face database, five of the Caltech 101 database and ten of Fifteen Scene Categories Dataset. Experimental results indicated that the method made full use of the distribution information of unlabeled image samples and made the image classification hyperplane more accurate.

Key words: image classification, unlabeled sample, probability density estimation, classification hyperplane, minimax probability machine

中图分类号: 

  • TP391
[1] 唐寿成. 图像分类方法的比较及应用[D].北京:北京邮电大学, 2011. TANG Shoucheng. Comparison of image classification method and its application[D].Beijing:Beijing University of Posts and Telecommunications, 2011.
[2] 朱建华, 刘政凯, 俞能海. 一种多光谱遥感图象的自适应最小距离分类方法[J].中国图象图形学报,2000, 5(1):21-24. ZHU Jianhua, LIU Zhengkai, YU Nenghai. Remote sensing image classification using an adaptive min-distance algorithm[J]. Journal of Image and Graphics, 2000, 5(1):21-24.
[3] HARTIGAN J A, WONG M A. Algorithm AS 136: a K-means clustering algorithm[J]. Applied Statistics, 1979, 29(1):100-108.
[4] WANG J. Locality-constrained linear coding for image classification[J]. IEEE Conference on Computer Vision & Pattern Recognition, 2010, 119(5):3360-3367.
[5] PRIYA T, PRASAD S, WU H. Superpixels for spatially reinforced bayesian classification of hyperspectral images[J]. IEEE Geoscience & Remote Sensing Letters, 2015, 12(5):1071-1075.
[6] BERTHOD M, KATO Z, SHAN Y, et al. Bayesian image classification using Markov random fields[J]. Image & Vision Computing, 1996, 14(4):285-295.
[7] NIGSCH F, BENDER A, VAN BUUREN B, et al. Melting point prediction employing k-nearest neighbor algorithms and genetic parameter optimization[J]. Journal of Chemical Information and Modeling, 2006, 46(6):2412-2422.
[8] HOULE M E, MA X, ORIA V, et al. Improving the quality of K-NN graphs through vector sparsification:application to image databases[J]. International Journal of Multimedia Information Retrieval, 2014, 3(4):259-274.
[9] SUYKENS J A K, VANDEWALLE J. Least squares support vector machine classifiers[J]. Neural Processing Letters, 1999, 9(3):293-300.
[10] LANCKRIET G R G, GHAOUI L E, BHATTACHARYYA C, et al. Minimax probability machine[J]. Advances in Neural Information Processing Systems, 2001, 37(1):192-200.
[11] WANG Y, HAN J. Study on iris image classification approach based on minimax probability machine[J]. Journal of Xi'an Jiaotong University, 2006, 40(6):643-651.
[12] NG J K C, ZHONG Y, YANG S. A comparative study of minimax probability machine-based approaches for face recognition[J]. Pattern Recognition Letters, 2007, 28(15):1995-2002.
[13] LANCKRIET G R G, EL GHAOUI L, BHATTACHARYYA C, et al. A robust minimax approach to classification[J]. Journal of Machine Learning Research, 2003, 3(3):555-582.
[14] BERTSIMAS D, POPESCU I. Optimal inequalities in probability theory: a convex optimization approach[J]. SIAM Journal on Optimization, 2005, 15(3):780-804.
[15] BERTSIMAS D, TSITSIKLIS J. Introduction to linear optimization[J]. Siam Review, 1998, 20(256):554-575.
[16] CHERNOFF H. The selection of effective attributes for deciding between hypotheses using linear discriminant functions[R]. California, USA:Univ Calif Dept of Statistics, 1970:56-60.
[17] 蒋艳凰, 赵強利. 机器学习方法[M]. 北京:电子工业出版社, 2009:5-30.
[18] 陈宝林. 最优化理论与算法[M]. 北京:清华大学出版社有限公司, 2005:281-322.
[19] XU J. An extended one-versus-rest support vector machine for multi-label classification[J]. Neurocomputing, 2011, 74(17):3114-3124.
[20] HSU C W, LIN C J. A comparison of methods for multiclass support vector machines[J]. Neural Networks, IEEE Transactions on, 2002, 13(2):415-425.
[21] (¨overU)BEYLI E D. ECG beats classification using multiclass support vector machines with error correcting output codes[J]. Digital Signal Processing, 2007, 17(3):675-684.
[1] 何其佳,刘振丙,徐涛,蒋淑洁. 基于LBP和极限学习机的脑部MR图像分类[J]. 山东大学学报(工学版), 2017, 47(2): 86-93.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!