• •    下一篇

基于超像素合并的高光谱图像分类

谢福鼎1, 李旭1, 黄丹2, 金翠1   

  1. 1. 辽宁师范大学地理科学学院, 大连 116029;
    2. 辽宁师范 大学计算机与信息技术学院, 大连 116081
  • 收稿日期:2021-07-15 修回日期:2021-11-02 出版日期:2021-12-25 发布日期:2022-03-16
  • 通讯作者: 金翠,Email:cuijin@lnnu.edu.cn
  • 基金资助:
    国家自然科学基金(41801340),辽宁省教育厅自然科学研究项目(LJ2019013)资助课题.

谢福鼎, 李旭, 黄丹, 金翠. 基于超像素合并的高光谱图像分类[J]. 系统科学与数学, 2021, 41(12): 3268-3279.

XIE Fuding, LI Xu, HUAGN Dan, JIN Cui. Superpixel Merging-Based Hyperspectral Image Classification[J]. Journal of Systems Science and Mathematical Sciences, 2021, 41(12): 3268-3279.

Superpixel Merging-Based Hyperspectral Image Classification

XIE Fuding1, LI Xu1, HUAGN Dan2, JIN Cui1   

  1. 1. Department of Geography, Liaoning Normal University, Dalian 116029;
    2. Department of Computer Science and Information Technology, Liaoning Normal University, Dalian 116081
  • Received:2021-07-15 Revised:2021-11-02 Online:2021-12-25 Published:2022-03-16
超像素级的高光谱图像分类是一类有代表性的谱-空分类方法.与像素级分类方法相比,超像素级的分类方法在分类精度和分类效率方面都有明显的优势.然而,超像素级分类算法的主要缺点是分类结果严重依赖于超像素的分割尺度.已有的工作表明,最优超像素分割尺度的获得往往是一个实验结果,很难预先确定.为了削弱这种依赖性,文章提出了一种基于超像素合并的超像素级高光谱分类算法.该方法首先采用局部模块度函数对所构造的稀疏加权超像素图进行合并;然后通过新定义的映射将每一个超像素块表示为一个样本点,使用流行的KNN方法对合并后的超像素图像进行超像素级分类.超像素的合并增强了空间信息在分类中的作用,有效地削弱了分类结果对超像素分割尺度的依赖性,并提高了分类精度.为了评价该方法的有效性,在4个公开的实际高光谱数据集上,将所提出的方法与一些竞争性的高光谱图像分类方法进行了实验和对比.实验结果和比较结果表明,该方法不仅有效削弱了超像素分割尺度对分类结果的影响,且在分类精度和计算效率方面都有十分明显的优势.
Superpixel-level hyperspectral image classification is a representative spec-tral-spatial classification method. Compared with the pixel-wise classification method, it has obvious advantages in classification accuracy and efficiency. However, the main disadvantage of superpixel-level classification algorithms is that the classification results depend heavily on the segmentation scale of superpixels. Existing literature shows that the optimal segmentation scale of superpixels is usually an experimental result, and it is difficult to be specified in advance. To weaken this dependency, a superpixel-level hyperspectral image classification algorithm based on superpixel merging is proposed in this work. Local modularity function is first used to merge the sparse weighted superpixel graph constructed. By the newly defined mapping, each superpixel is represented as a sample. Then popular KNN method is adopted to classify the merged image at the superpixel level. The superpixel merging enhances the role of spatial information in classification, effectively weakens the dependence of classification results on the segmentation scale of superpixels, and improves the classification accuracy. To evaluate the effectiveness of the method, the proposed algorithm is compared with some competitive hyperspectral image classification methods on four publicly real hyperspectral datasets. The experimental and comparative results show that the proposed method not only effectively reduces the influence of superpixel segmentation scale on the classification results, but also has obvious advantages both in classification accuracy and computational efficiency.

MR(2010)主题分类: 

()
[1] Landgrebe D A, Serpico S, Crawford M, et al. Introduction to the special issue on analysis of hyperspectral image data. IEEE Trans. Geo. Remot. Sens., 2001, 39(7):1343-1345.
[2] Yuen P, Richardson M. An introduction to hyperspectral imaging and its application for security, surveillance and target acquisition. Imag. Sci. J., 2010, 58(5):241-253.
[3] Fauvel M, Tarabalka Y, Benediktsson J A, et al. Advances in spectral-spatial classification of hyperspectral images. Proc. IEEE, 2013, 101(3):652-675.
[4] Kang X, Li S, Benediktsson J A. Spectral-spatial hyperspectral image classification with edgepreserving filtering. IEEE Trans. Geo. Remot. Sens., 2014, 52(5):2666-2677.
[5] Kang X, Li S, Benediktsson J A. Feature extraction of hyperspectral images with image fusion and recursive filtering. IEEE Trans. Geo. Remot. Sens., 2014, 52(6):3742-3752.
[6] Fang L, Li S, Duan W, et al. Classification of hyperspectral images by exploiting spectral-spatial information of superpixel via multiple kernels. IEEE Trans. Geo. Remot. Sens., 2015, 53(12):6663-6674.
[7] Jiang J, Ma J, Chen C, et al. Super PCA:A superpixelwise PCA approach for unsupervised feature extraction of hyperspectral imagery. IEEE Trans. Geo. Remot. Sens., 2018, 56(8):4581-4593.
[8] Xie F, Lei C, Yang J, et al. An effective classification scheme for hyperspectral image based on superpixel and discontinuity preserving relaxation. Remot. Sens., 2019, 11(10):1149.
[9] Xie F, Lei C, Jin C, et al. A novel spectral-spatial classification method for hyperspectral image at superpixel level. Appl. Sci., 2020, 10(2):463.
[10] Beaulieu J, Goldberg M. Hierarchy in picture segmentation:A stepwise optimization approach. IEEE Trans. Patt. Anal. Mach. Intel., 1989, 11(2):150-163.
[11] Tilton J. Image segmentation by region growing and spectral clustering with a natural convergence criterion. IEEE Inter. Geo. Remot. Sens. Symp., 1998, 4:1766-1768.
[12] Haris K, Efstratiadis S, Maglaveras N, et al. Hybrid image segmentation using watersheds and fast region merging. IEEE Trans. Imag. Proc., 1998, 7(12):1684-1699.
[13] Murtagh F, Contreras P. Algorithms for hierarchical clustering:An overview. Wiley. Interdiscip. Rev. Data. Min. Knowl. Discov., 2012, 2(1):86-97.
[14] Kennedy S M, Williamson W, Roth J D, et al. Cluster-based spectral-spatial segmentation of hyperspectral imagery. IEEE Access, 2020, 8:140361-140391.
[15] Johnson B, Xie Z. Unsupervised image segmentation evaluation and refinement using a multiscale approach. ISPRS J. Photo. Remot. Sens., 2011, 66(4):473-483.
[16] Lu T, Li S, Fang L, et al. Set-to-set distance-based spectral-spatial classification of hyperspectral images. IEEE Trans. Geo. Remot. Sens, 2016, 54(12):7122-7134.
[17] Zhao Y, Su F, Yan F. Novel semi-supervised hyperspectral image classification based on a superpixel graph and discrete potential method. Remot. Sens., 2020, 12(9):1528.
[18] Tu B, Wang J, Kang X, et al. KNN-based representation of superpixels for hyperspectral image classification. IEEE J. Sel. Top. Appl. Ear. Obs. Remot. Sens., 2018, 11(11):4032-4047.
[19] Achanta R, Shaji A, Smith K, et al. SLIC superpixels compared to state-of-the-art superpixel methods. IEEE Trans. Patt. Ana. Mach. Intell., 2012, 34(11):2274-2282.
[20] Liu M, Tuzel O, Ramalingam S, et al. Entropy rate superpixel segmentation. CVPR, 2011, 2097-2104.
[21] Vincent L, Soille P. Watersheds in digital spaces:An efficient algorithm based on immersion simulations. IEEE Trans. Patt. Ana. Mach. Intell., 1991, 13(6):583-598.
[22] Jianbo S, Jitendra M. Normalized cuts and image segmentation. IEEE Trans. Patt. Ana. Mach. Intell., 2000, 22(8):888-905.
[23] Newman M E J, Girvan M. Finding and evaluating community structure in networks. Phys. Rev. E, 2004, 69(2):026113-1-0261131-16.
[24] Clauset A. Finding local community structure in networks. Phys. Rev. E, 2005, 72(2):026132.
[25] Ward J H. Hierarchical grouping to optimize an objective function. Cogn. Ling., 1963, 58(301):236-244.
[26] Szekely G, Rizzo M. Hierarchical clustering via joint between-within distances:Extending ward's minimum variance method. J. Class., 2005, 22(2):151-183.
[27] Sun L, Wu Z, Liu J, et al. Supervised spectral-spatial hyperspectral image classification with weighted Markov random fields. IEEE Trans. Geo. Remot. Sens., 2015, 53(3):1490-1503.
[28] Zhong Y, Hu X, Luo C, et al. WHU-Hi:UAV-borne hyperspectral with high spatial resolution (H2) benchmark datasets and classifier for precise crop identification based on deep convolutional neural network with CRF. Remote Sens. Environ., 2020, 250:112012.
[1] 琚春华,陈冠宇,鲍福光. 基于kNN-Smote-LSTM的消费金融风险检测模型[J]. 系统科学与数学, 2021, 41(2): 481-498.
[2] 叶菲菲,杨隆浩,王应明. 区域环境污染强度测算及其分类治理效率评价研究[J]. 系统科学与数学, 2020, 40(6): 984-1003.
[3] 古楠楠,孙湘南,刘伟,李路云. 基于自步学习与稀疏自表达的半监督分类方法[J]. 系统科学与数学, 2020, 40(1): 191-208.
[4] 贾玉改,唐锡晋. 在线极化辩论立场与理性表达关系研究[J]. 系统科学与数学, 2019, 39(10): 1642-1654.
[5] 谢福鼎,雷存款,李芳菲,嵇敏. 基于模糊c均值算法和人工蜂群算法的无监督波段选择[J]. 系统科学与数学, 2018, 38(12): 1417-1428.
[6] 姚娆,谢福鼎. 基于谱-空-纹特征融合的高光谱影像分类方法[J]. 系统科学与数学, 2017, 37(10): 2103-2110.
[7] 张少白,曾又,刘友谊. 基于DIVA模型的脑电信号识别方法[J]. 系统科学与数学, 2015, 35(5): 489-498.
[8] 冯光庭,殷秀芝,凌中华,张兴安. 一类~${\bm n}{\bf +1}$~次平面拟齐次向量场的全局性质[J]. 系统科学与数学, 2014, 34(4): 475-487.
[9] 李为相,李帮义. 一种基于支持向量域描述的区间数分类[J]. 系统科学与数学, 2012, 32(3): 319-326.
[10] 周丽, 朱杰, 郭键. 分类存储返回型和S型拣选路径随机模型的比较研究[J]. 系统科学与数学, 2011, 31(8): 921-931.
[11] 俞建. n人非合作博弈的轻微利他平衡点[J]. 系统科学与数学, 2011, 31(5): 534-539.
[12] 特木尔朝鲁;张智勇. 一类非线性波动方程的势对称分类[J]. 系统科学与数学, 2009, 29(3): 389-411.
[13] 卢志义;刘乐平;孟生旺. 基于污染Gamma分布的聚合风险模型及其在风险分类中的应用[J]. 系统科学与数学, 2009, 29(2): 174-183.
[14] 余乐安;汪寿阳 . 基于核主元分析的带可变惩罚因子最小二乘模糊支持向量机模型及其在信用分类中的应用[J]. 系统科学与数学, 2009, 29(10): 1311-1326.
[15] 张智勇;雍雪林;陈玉福. 一类组合方程的势对称分类[J]. 系统科学与数学, 2008, 28(8): 905-914.
阅读次数
全文


摘要