基于深度特征分析的双线性图像相似度匹配算法

doi:10.11772/j.issn.1001-9081.2016.10.2822

计算机应用 ›› 2016, Vol. 36 ›› Issue (10): 2822-2825.DOI: 10.11772/j.issn.1001-9081.2016.10.2822

基于深度特征分析的双线性图像相似度匹配算法

李鸣^1,2, 张鸿^1,2

1. 武汉科技大学计算机科学与技术学院, 武汉 430065;
2. 智能信息处理与实时工业系统湖北省重点实验室(武汉科技大学), 武汉 430065

收稿日期:2016-03-23 修回日期:2016-06-18 发布日期:2016-10-10
通讯作者: 张鸿,E-mail:zhanghong_wust@163.com
作者简介:李鸣(1992—),男,湖北武汉人,硕士研究生,主要研究方向:图像检索、深度学习;张鸿(1979—),女,湖北襄阳人,教授,博士,CCF会员,主要研究方向:基于内容的多媒体检索、数据挖掘、机器学习。
基金资助:
国家自然科学基金资助项目（61373109，61003127）。

Bilinear image similarity matching algorithm based on deep feature analysis

LI Ming^1,2, ZHANG Hong^1,2

1. College of Computer Science and Technology, Wuhan University of Science and Technology, Wuhan Hubei 430065, China;
2. Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial Systems (Wuhan University of Science and Technology), Wuhan Hubei 430065, China

Received:2016-03-23 Revised:2016-06-18 Published:2016-10-10
Supported by:
BackgroundThis work is supported by the National Natural Science Foundation of China (61373109,61003127).

摘要/Abstract

摘要： 基于内容的图像检索一直面临"语义鸿沟"的难题，特征选择对语义学习结果有着直接的影响；而传统距离度量方法往往从单一角度进行相似性计算，不能很好地表示出图像之间的相似度。为了解决以上问题，提出基于深度特征分析的双线性图像相似度匹配的方法。首先，将图像数据集在卷积神经网络模型上进行微调训练，然后利用训练好的卷积神经网络对图像进行特征提取，获得全连接层输出的特征之后，通过双线性相似性度量方法得到图像间相似度的大小，通过对相似度的大小排序，返回最相似的图像实例。在Caltech101和Caltech256数据集上的对比实验显示，所提算法的平均查准率、TopK查准率和查全率均优于对比算法，验证了所提算法的有效性。

关键词: 深度神经网络, 双线性相似度, 图像检索, 语义鸿沟, 平均查准率

Abstract: Content-based image retrieval has being faced the problem of "semantic gap", feature selection has a direct influence on semantic learning results; while traditional distance metric often calculates the similarity from a single perspective, which cannot well express the similarity between images. To resolve the above problem, a bilinear image similarity matching algorithm based on deep feature analysis was proposed. First, the image dataset was fine-tuning trained on the Convolutional Neural Network (CNN) model, then the image features were extracted by using the trained CNN. After getting the output features of the full connection layer, the image similarity was calculated by the bilinear similarity matching algorithm, and the most similar image instance was returned after sorting the similarity. Experimental results on Caltech101 and Caltech 256 datasets show that compared with the contrast algorithms, the proposed algorithm can get higher mean average precision, TopK precision and recall, which demonstrates the effectiveness of the proposed algorithm.

Key words: deep neural network, bilinear image similarity matching, image retrieval, semantic gap, mean average precision

中图分类号:

TP391.413

李鸣, 张鸿. 基于深度特征分析的双线性图像相似度匹配算法[J]. 计算机应用, 2016, 36(10): 2822-2825.

LI Ming, ZHANG Hong. Bilinear image similarity matching algorithm based on deep feature analysis[J]. Journal of Computer Applications, 2016, 36(10): 2822-2825.

参考文献

[1] LOWE D G. Distinctive image features from scale-invariant key-points[J]. International Journal of Computer Vision, 2004, 60(2):91-110.
[2] 何云峰, 周玲, 于俊清, 等. 基于局部特征聚合的图像检索方法[J]. 计算机学报, 2011, 34(11):2224-2233.(HE Y F, ZHOU L, YU J Q, et al. Image retrieval based on locally features aggregating[J]. Chinese Journal of Computers, 2011, 34(11):2224-2233.)
[3] BENGIO Y, COURVILLE A, VINCENT P. Representation learning: a review and new perspectives[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2013, 35(8):1798-828.
[4] 张鸿, 吴飞, 张晓龙. 基于关系矩阵融合的多媒体数据聚类[J]. 计算机学报, 2011, 34(9):1705-1711.(ZHANG H, WU F, ZHANG X L. Multimedia data clustering based on correlation matrix fusion [J]. Chinese Journal of Computers, 2011, 34(9):1705-1711.)
[5] ZHANG H, GAO X, WU P, et al. A cross-media distance metric learning framework based on multi-view correlation mining and matching[J]. World Wide Web, 2016,19(2):181-197.
[6] BECKER B C, ORTIZ E G. Evaluating open-universe face identification on the Web[C]//Proceedings of the 2013 IEEE Conference on Computer Vision and Pattern Recognition Workshops. Washington, DC: IEEE Computer Society, 2013:904-911.
[7] DENTON E, ZAREMBA W, BRUNA J, et al. Exploiting linear structure within convolutional networks for efficient evaluation[EB/OL].[2015-10-10]. machinelearning.wustl.edu/mlpapers/paper_files/NIPS2014_5544.pdf.
[8] HINTON G, DENG L, YU D, et al. Deep neural networks for acoustic modeling in speech recognition: the shared views of four research groups[J]. IEEE Signal Processing Magazine, 2012, 29(6): 82-97.
[9] SIMARD P Y, STEINKRAUS D, PLATT J C. Best practices for convolutional neural networks applied to visual document analysis[C]//Proceedings of the 2003 International Conference on Document Analysis and Recognition. Washington, DC: IEEE Computer Society, 2003:958.
[10] KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012, 25(2):2012.
[11] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, NJ: IEEE, 2014:580-587.
[12] SIMON M, RODNER E, DENZLER J. Part detector discovery in deep convolutional neural networks[C]//Proceedings of the 12th Asian Conference on Computer Vision, LNCS 9004. Piscataway, NJ: IEEE, 2014:162-177.
[13] WEINBERGER K Q, SAUL L K. Fast solvers and efficient implementations for distance metric learning[C]//Proceedings of the 25th International Conference on Machine Learning. New York: ACM, 2008: 1160-1167.
[14] CHECHIK G, SHARMA V, SHALIT U, et al. Large scale online learning of image similarity through ranking[J]. The Journal of Machine Learning Research, 2010, 11: 1109-1135.
[15] CRAMMER K, DEKEL O, KESHET J, et al. Online passive-aggressive algorithms[J]. The Journal of Machine Learning Research, 2006, 7: 551-585.
[16] LI F F, FERGUS R, PERONA P. Learning generative visual models from few training examples: an incremental Bayesian approach tested on 101 object categories[J]. Computer Vision & Image Understanding, 2007, 106(1):59-70.
[17] ZHENG W S, GONG S, XIANG T. Unsupervised selective transfer learning for object recognition[C]//Proceedings of the 2010 Asian Conference on Computer Vision. Piscataway, NJ: IEEE, 2010:527-541.
[18] DUVENAUD D, RIPPEL O, ADAMS R P, et al. Avoiding pathologies in very deep networks[EB/OL]. [2015-01-01]. http://arxiv.org/abs/1402.5836.
[19] LI F F, PERONA P. A Bayesian hierarchical model for learning natural scene categories[C]//Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision & Pattern Recognition. Washington, DC: IEEE Computer Society, 2005:524-531.
[20] 顾昕, 张兴亮, 王超, 等. 基于文本和内容的图像检索算法[J]. 计算机应用, 2014, 34(增刊2):280-282.(GU X, ZHANG X L, WANG C, et al. Image retrieval algorithm based on text and content[J]. Journal of Computer Applications, 2014, 34(S2):280-282.)

基于深度特征分析的双线性图像相似度匹配算法

Bilinear image similarity matching algorithm based on deep feature analysis

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

[1]	杨粟, 欧阳智, 杜逆索. 基于相关度距离的无监督并行哈希图像检索[J]. 计算机应用, 2021, 41(7): 1902-1907.
[2]	王曙燕, 侯则昱, 孙家泽. 面向深度学习的对抗样本差异性检测方法[J]. 计算机应用, 2021, 41(7): 1849-1856.
[3]	张明明, 卢庆宁, 李文中, 宋浒. 基于联合动态剪枝的深度神经网络压缩算法[J]. 计算机应用, 2021, 41(6): 1589-1596.
[4]	张文烨, 尚方信, 郭浩. 基于Octave卷积的混合精度神经网络量化方法[J]. 计算机应用, 2021, 41(5): 1299-1304.
[5]	杨丽, 王时绘, 朱博. 基于动态和静态偏好的兴趣点推荐算法[J]. 计算机应用, 2021, 41(2): 398-406.
[6]	陆荣秀, 陈明明, 杨辉, 朱建勇. 基于溶液图像时序特征的元素组分含量动态监测系统[J]. 计算机应用, 2021, 41(10): 3075-3081.
[7]	陈彦如, 张涂静娃, 杜千, 冉茂亮, 王红军. 基于深度森林的高铁站室内热舒适度等级预测[J]. 计算机应用, 2021, 41(1): 258-264.
[8]	李鸣, 郭晨皓, 陈星. 视觉类深度神经网络的自动标注[J]. 计算机应用, 2020, 40(6): 1593-1600.
[9]	韦伟, 李小娟. 基于相似论文增广的深度学习专利质量评估[J]. 计算机应用, 2020, 40(4): 966-971.
[10]	樊琦, 李卓, 陈昕. 基于边缘计算的分支神经网络模型推断延迟优化[J]. 计算机应用, 2020, 40(2): 342-346.
[11]	邓凯, 黄佳进, 秦进. 基于物品的统一推荐模型[J]. 计算机应用, 2020, 40(2): 530-534.
[12]	顾军华, 王锋, 戚永军, 孙哲然, 田泽培, 张亚娟. 基于多尺度卷积特征融合的肺结节图像检索方法[J]. 计算机应用, 2020, 40(2): 561-565.
[13]	杨坚伟, 严群, 姚剑敏, 林志贤. 基于深度神经网络的移动端人像分割[J]. 计算机应用, 2020, 40(12): 3644-3650.
[14]	杨磊, 赵红东. 基于轻量级深度神经网络的环境声音识别[J]. 计算机应用, 2020, 40(11): 3172-3177.
[15]	朱倩倩, 刘渊, 李甫. 深度神经网络的仿生矩阵约简与量化方法[J]. 计算机应用, 2020, 40(10): 2817-2821.