[1]鲍文霞,余国芬,朱明,等.基于椭圆形度量谱特征的图像匹配算法[J].东南大学学报(自然科学版),2018,48(3):387-392.[doi:10.3969/j.issn.1001-0505.2018.03.002]
 Bao Wenxia,Yu Guofen,Zhu Ming,et al.Image matching algorithm based on elliptic metric spectral feature[J].Journal of Southeast University (Natural Science Edition),2018,48(3):387-392.[doi:10.3969/j.issn.1001-0505.2018.03.002]
点击复制

基于椭圆形度量谱特征的图像匹配算法()
分享到:

《东南大学学报(自然科学版)》[ISSN:1001-0505/CN:32-1178/N]

卷:
48
期数:
2018年第3期
页码:
387-392
栏目:
计算机科学与工程
出版日期:
2018-05-20

文章信息/Info

Title:
Image matching algorithm based on elliptic metric spectral feature
作者:
鲍文霞12余国芬1朱明1梁栋1
1安徽大学计算智能与信号处理教育部重点实验室, 合肥 230039 ; 2中国人民解放军陆军军官学院偏振光成像探测技术安徽省重点实验室, 合肥 230031
Author(s):
Bao Wenxia12 Yu Guofen1 Zhu Ming1 Liang Dong1
1 Key Laboratory of Intelligent Computing and Signal Processing of Education of Ministry, Anhui University, Hefei 230039, China
2 Anhui Province Key Laboratory of Polarization Imaging Detection Technology, Chinese Peoples Liberation Army Academy, Hefei 230031, China
关键词:
图像匹配 椭圆形度量 谱特征 匹配数学模型 贪心算法
Keywords:
image matching elliptic metric spectral feature matching mathematical model greedy algorithm
分类号:
TP391
DOI:
10.3969/j.issn.1001-0505.2018.03.002
摘要:
为了进一步提高基于谱特征的图像匹配算法的精度并拓宽其应用范围,提出了一种基于椭圆形度量谱特征的图像匹配算法.通过引入对样本数据具有更好区分性的椭圆形几何,结合数据统计特性定义了椭圆形度量.对特征点构造谱特征时,根据椭圆形相对距离选择子特征点集并构造无向加权图,对利用椭圆形度量获取的关联邻接矩阵进行谱分解,基于特征值和谱隙向量的统计量构造椭圆形度量谱特征.在特征点匹配过程中,根据椭圆形距离度量谱特征之间的相似性,建立匹配数学模型,并采用贪心算法进行求解.针对序列图像以及视角变换、形变较大图像的对比实验结果表明,所提算法的匹配正确率保持100%,优于其他谱特征匹配算法.椭圆形度量谱特征提高了匹配算法的精度,对噪声具有较高的鲁棒性.
Abstract:
In order to further improve the accuracy of the image matching algorithm based on spectral features and broaden the application range, an image matching algorithm based on the elliptic metric spectral feature is proposed. By introducing elliptic geometry with better distinguishability of sample data, an elliptic measure is defined based on the statistical properties of the data. When constructing the spectral features of a feature point, the sub feature points are selected according to the elliptic distance and the undirected weighted graph is constructed. The spectral decomposition of the associated adjacency matrix obtained by the elliptic metric is carried out, and the elliptic metric spectral feature is constructed by the statistics of the eigenvalues and the spectral gap vectors. During the process of feature point matching, the elliptic distance is used to measure the similarity between spectral features, and a matching mathematical model is established and solved by the greedy algorithm. The contrast experimental results on sequence images, visual angle transformation and large deformed images show that the matching accuracy of the proposed algorithm keeps 100%, which is superior to other spectral feature matching algorithms. The elliptic metric spectral feature can improve the accuracy of the matching algorithm and is robust to noise.

参考文献/References:

[1] Carcassoni M, Hancock E R. Spectral correspondence for point pattern matching[J]. Pattern Recognition, 2003, 36(1): 193-204. DOI:10.1016/s0031-3203(02)00054-7.
[2] Feng W, Liu Z Q, Wan L, et al. A spectral-multiplicity-tolerant approach to robust graph matching[J]. Pattern Recognition, 2013, 46(10): 2819-2829. DOI:10.1016/j.patcog.2013.03.003.
[3] Yang X, Qiao H, Liu Z Y. Feature correspondence based on directed structural model matching[J]. Image and Vision Computing, 2015, 33: 57-67. DOI:10.1016/j.imavis.2014.11.001.
[4] Peng X, Yu Z, Yi Z, et al. Constructing the L2-graph for robust subspace learning and subspace clustering[J]. IEEE Trans Cybern, 2017, 47(4): 1053-1066. DOI:10.1109/TCYB.2016.2536752.
[5] Leordeanu M, Hebert M. A spectral technique for correspondence problems using pairwise constraints [C]//Tenth IEEE International Conference on Computer Vision. Beijing, China, 2005: 1482-1489. DOI:10.1109/iccv.2005.20.
[6] Yuan Y, Pang Y, Wang K, et al. Efficient image matching using weighted voting[J]. Pattern Recognition Letters, 2012, 33(4): 471-475. DOI:10.1016/j.patrec.2011.02.008.
[7] Tang J, Shao L, Zhen X. Robust point pattern matching based on spectral context[J]. Pattern Recognition, 2014, 47(3): 1469-1484. DOI:10.1016/j.patcog.2013.09.017.
[8] 朱明,梁栋,范益政,等.基于谱特征的图像匹配算法 [J].华南理工大学学报(自然科学版),2015,43(9): 60-66. DOI:10.3969/j.issn.1000-565X.2015.09.010.
Zhu Ming, Liang Dong, Fan Yizheng, et al. Image matching algorithm based on spectral feature [J]. Journal of South China University of Technology(Natural Science Edition), 2015, 43(9): 60-66. DOI:10.3969/j.issn.1000-565X.2015.09.010. (in Chinese)
[9] Yan P, Liang D, Tang J, et al. Local feature descriptor invariant to monotonic illumination changes[J]. Journal of Electronic Imaging, 2016, 25(1): 013023. DOI:10.1117/1.jei.25.1.013023.
[10] Bo D, Zhang G L, Cui X L. An algorithm of image matching based on Mahalanobis distance and weighted KNN graph [C]//2015 International Conference on Information Science and Control Engineering. Shanghai, China, 2015: 116-121.
[11] Famouri M, Azimifar Z. A statistical approach to rank the matched image points [C]//2016 24th Iranian Conference on Electrical Engineering. Shiraz, Iran, 2016: 1214-1218. DOI:10.1109/iraniancee.2016.7585706.
[12] Klein F. Ueber die sogenannte Nicht-Euklidische geometrie[J]. Mathematische Annalen, 1970, 6(2): 112-145. DOI:10.1007/BF01443189.
[13] Zhou F, Torre F D L. Deformable Graph Matching [C]//2013 IEEE Conference on Computer Vision and Pattern Recognition. Portland, USA, 2013: 2922-2929. DOI:10.1109/cvpr.2013.376.
[14] Cho M, Sun J, Duchenne O, et al. Finding matches in a Haystack: A max-pooling strategy for graph matching in the presence of outliers [C]//Computer Vision and Pattern Recognition. Columbus, OH, USA, 2014: 2091-2098.DOI:10.1109/cvpr.2014.268.

备注/Memo

备注/Memo:
收稿日期: 2017-12-04.
作者简介: 鲍文霞(1980—),女,博士,副教授,bwxia@ahu.edu.cn.
基金项目: 国家自然科学基金资助项目(61401001,61501003,61672032)、中国人民解放军陆军军官学院偏振光成像探测技术安徽省重点实验室开放基金项目资助(2016-KFJJ-001).
引用本文: 鲍文霞,余国芬,朱明,等.基于椭圆形度量谱特征的图像匹配算法[J].东南大学学报(自然科学版),2018,48(3):387-392. DOI:10.3969/j.issn.1001-0505.2018.03.002.
更新日期/Last Update: 2018-05-20