Please wait a minute...
 
自然资源遥感  2022, Vol. 34 Issue (2): 37-46    DOI: 10.6046/zrzyyg.2021214
  技术方法 本期目录 | 过刊浏览 | 高级检索 |
基于空间统计学的高光谱遥感影像主成分选择方法
孙肖1(), 彭军还2(), 赵锋3, 王晓阳1, 吕洁2, 张登峰4
1.中国地质调查局廊坊自然资源综合调查中心,廊坊 065000
2.中国地质大学(北京)土地科学技术学院,北京 100083
3.中国地质调查局乌鲁木齐自然资源综合调查中心, 乌鲁木齐 830057
4.中国地质调查局西安矿产资源调查中心,西安 710100
Principal component selection method for hyperspectral remote sensing images based on spatial statistics
SUN Xiao1(), PENG Junhuan2(), ZHAO Feng3, WANG Xiaoyang1, LYU Jie2, ZHANG Dengfeng4
1. Langfang Natural Resources Comprehensive Survey Center, China Geological Survey, Langfang 065000, China
2. School of Land Science and Technology, China University of Geosciences (Beijing), Beijing 100083, China
3. Urumqi Natural Resources Comprehensive Survey Center, China Geological Survey, Urumqi 830057, China
4. Xi’an Center of Mineral Resources Survey, China Geological Survey, Xi’an 710100, China
全文: PDF(10623 KB)   HTML  
输出: BibTeX | EndNote (RIS)      
摘要 

主成分分析是一种广泛使用的高光谱遥感影像降维方法,在面向任务的工作中,基于累计方差贡献率的主成分选择方法效果并不理想。针对主成分分析变换后主成分选择的问题,提出基于空间统计学的主成分选择方法。计算各主成分的半变异函数参数变程、拱高、基台值,综合变程和拱高/基台值实现主成分的选择。变程的大小用以判断每一个主成分空间相关性的范围,拱高/基台值的大小用以判断每一个主成分空间相关性的强弱。仿真实验证明了变程和拱高/基台值可以有效表达高光谱遥感影像空间相关性的范围和强弱。在真实高光谱遥感影像实验的基础上,从主观和客观2个方面来综合确定主成分选择的经验阈值,即变程为2.5、拱高/基台值为0.2。从基于支持向量机算法的分类结果来看,和传统方法相比,利用变程和拱高/基台值可以筛选出图像质量较好的主成分,不仅能够达到降维的目的,同时能够保证足够高的分类精度。

服务
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章
孙肖
彭军还
赵锋
王晓阳
吕洁
张登峰
关键词 高光谱主成分分析空间统计学半变异函数支持向量机    
Abstract

The principal component analysis is a widely used method for dimensionality reduction of hyperspectral remote sensing images. In task-oriented work, the principal component selection method based on cumulative variance contribution rate is not ideal. To address the problem of principal component selection after principal component analysis transformation, a method of principal component selection based on spatial statistics is proposed. The selection of principal components is performed by calculating the values of the semi-variogram parameter range and partial sill/sill of each principal component. The magnitude of a range is used to judge the range of spatial correlation of each principal component, and the partial sill/sill is used to judge the strength of spatial correlation of each principal component. The simulation proves that the variable range and partial sill/sill can effectively express the range and strength of spatial correlation of hyperspectral remote sensing images. Based on the experiment of real hyperspectral remote sensing images, the empirical threshold of principal component selection is determined from subjective and objective aspects, that is, the range is 2.5, and the partial sill/sill is 0.2. According to the classification results based on the support vector machine algorithm, compared with traditional methods, the principal components with better image quality can be screened by using variable range and partial sill/sill, which can not only achieve the purpose of dimensionality reduction, but also ensure high classification accuracy.

Key wordshyperspectral    principal component analysis    spatial statistics    semi-variogram    support vector machine
收稿日期: 2021-07-14      出版日期: 2022-06-20
ZTFLH:  P962  
基金资助:中国地质调查局项目“京津冀廊坊地区生态修复支撑调查”(DD20208073)
通讯作者: 彭军还
作者简介: 孙 肖(1988-),男,硕士,助理工程师,主要从事高光谱遥感解译研究。Email: sunxiao@mail.cgs.gov.cn
引用本文:   
孙肖, 彭军还, 赵锋, 王晓阳, 吕洁, 张登峰. 基于空间统计学的高光谱遥感影像主成分选择方法[J]. 自然资源遥感, 2022, 34(2): 37-46.
SUN Xiao, PENG Junhuan, ZHAO Feng, WANG Xiaoyang, LYU Jie, ZHANG Dengfeng. Principal component selection method for hyperspectral remote sensing images based on spatial statistics. Remote Sensing for Natural Resources, 2022, 34(2): 37-46.
链接本文:  
https://www.gtzyyg.com/CN/10.6046/zrzyyg.2021214      或      https://www.gtzyyg.com/CN/Y2022/V34/I2/37
Fig.1  算法流程图
Fig.2  仿真图像(不同栅格大小W=1,3,…,23; 信噪比SNR=10,20,30,45)
Fig.3  不同栅格大小仿真图像计算的变程结果
Fig.4  不同栅格大小仿真图像计算的拱高/基台值结果
Fig.5  Indian Pines数据集
Fig.6  Pavia U数据集
Fig.7  Salinas数据集
Fig.8  真实数据半变异函数参数计算结果图
Fig.9  Indian Pines数据集PCA后各主成分缩略图
不同方法 筛选的主成分编号
PC(2~0.25) 1~8,10~21,23~25,28,38,66,68,88,92,101,108,109
PC(2~0.2) 1~8,10~25,28,29,38,55,59,66,88,92,101,103,108,109
PC(2.5~0.25) 1~8,10~21,23~25,28,38,66,68,88,92,101
PC(2.5~0.2) 1~8,10~25,28,29,38,59,66,68,88,92,101,103
Tab.1  Indian Pines数据集不同阈值筛选的主成分
数据集 PC
(2~0.25)
PC
(2~0.2)
PC
(2.5~0.25)
PC
(2.5~0.2)
Indian Pines 0.912 0.912 0.905 0.906
Pavia U 0.856 0.881 0.858 0.883
Salinas 0.953 0.953 0.954 0.953
Tab.2  不同阈值的Kappa系数
Fig.10  Indian Pines,Pavia U和Salinas数据集分类结果图
数据集 PC
(0.99)
PC
(10%)
PC
(Entropy)
PC
(2.5~0.2)
Indian Pines 0.864 0.853 0.899 0.906
Pavia U 0.718 0.884 0.884 0.883
Salinas 0.910 0.952 0.951 0.953
Tab.3  不同方法Kappa系数
Fig.11  Indian Pines,Pavia U和Salinas数据集分类结果图
地物种类 PC
(0.99)
PC
(10%)
PC
(Entropy)
PC
(2.5~0.2)
塔楼 100 100 100 100
建筑物 74.02 75.2 81.45 79.53
林地 98.89 98.89 99.43 99.51
小麦地 100 100 100 100
大豆地 84.46 79.77 98.4 97.95
大豆略耕地 86.78 85.56 90.82 92.7
大豆未耕地 89.4 89.18 88.26 86.98
燕麦地 77.78 77.78 60 88.89
干草地 100 100 100 100
牧草已割地 81.82 90.91 94.12 100
草树地 100 99.77 100 100
牧草地 100 99.24 100 100
苜蓿地 85.71 80 86.84 85.71
玉米地 95.95 93.92 96.95 97.97
玉米未耕地 73.32 73.08 80.39 82.93
玉米略耕地 62.14 57.96 65.44 65.54
Tab.4  Indian Pines数据集制图精度
地物种类 PC
(0.99)
PC
(10%)
PC
(Entropy)
PC
(2.5~0.2)
草地 95.3 94.72 94.72 93.54
砂砾 42.87 72.63 72.63 72.00
金属板 100 99.91 99.91 99.82
裸地 31.51 79.64 79.64 81.02
柏油房顶 37.84 67.86 67.86 74.26
阴影 100 100 100 99.83
砖块 88.78 94.31 94.31 93.75
94.58 95.18 95.18 95.41
沥青路面 94.45 97.07 97.07 96.44
Tab.5  Pavia U数据集制图精度
地物种类 PC
(0.99)
PC
(10%)
PC
(Entropy)
PC
(2.5~0.2)
葡萄园垂直架子 98.23 99.92 99.84 99.83
未培育的葡萄园1 46.83 74.3 72.72 75.26
长叶莴苣_7wk 98.66 99.81 99.04 99.81
长叶莴苣_6wk 98.95 99.48 99.12 99.48
长叶莴苣_5wk 100 100 99.89 100
长叶莴苣_4wk 100 100 100 100
衰败的绿色杂草 99.35 100 100 100
生长中的葡萄园 100 100 100 100
未培育的葡萄园2 94.21 94.62 94.62 94.77
芹菜 100 100 100 100
作物残留 100 100 100 100
平整的休耕地 99.83 100 100 100
粗糙的休耕地 100 100 100 100
休耕地 100 100 100 100
椰菜2 99.95 100 100 100
椰菜1 99.49 99.91 100 100
Tab.6  Salinas数据集制图精度
数据集 不同方法 筛选的主成分编号 个数
Indian Pines PC(0.99) 1~25 25
PC(10%) 1~20 20
PC(Entropy) 1~46,49 47
PC(2.5~0.2) 1~8,10~25,28,29,38,59,66,68,88,92,101,103 34
Pavia U PC(0.99) 1~4 4
PC(10%) 1-10 10
PC(Entropy) 1-10 10
PC(2.5~0.2) 1~10,16,19,23,24,43,71 16
Salinas PC(0.99) 1~3 3
PC(10%) 1~20 20
PC(Entropy) 1~100 100
PC(2.5~0.2) 1~10,12,15~23,25,28,29,34,42,61,66,75,77,118,126,134 32
Tab.7  不同数据集筛选出的主成分信息
[1] Hughes G. On the mean accuracy of statistical pattern recognizers[J]. IEEE Transactions on Information Theory, 1968, 14(1):55-63.
doi: 10.1109/TIT.1968.1054102
[2] Chang C-I. Hyperspectral Data Exploitation:Theory and Applications[M]. John Wiley & Sons Inc,Hoboken,New Jersey, 2007,205.
[3] 宋海峰, 陈广胜, 杨巍巍. 基于PCA的高光谱遥感图像分类[J]. 测绘工程, 2017, 26(12):17-20,26.
Song H F, Chen G S, Yang W W. Principal component analysis for hyper spectral image classification[J]. Engineering of Surveying and Mapping, 2017, 26(12):17-20,26.
[4] Jolliffe I T. Principal Component Analysis[M]. Wiley, 2nd edition, 2002.
[5] Pu R, Gong P, Tian Y, et al. Invasive species change detection using artificial neural networks and CASI hyperspectral imagery[J]. Environmental Monitoring and Assessment, 2008, 140(1-3):15-32.
doi: 10.1007/s10661-007-9843-7
[6] Chen G, Qian S. Denoising and dimensionality reduction of hyperspectral imagery using wavelet packets,neighbour shrinking and principal component analysis[J]. International Journal of Remote Sensing, 2009, 30(18):4889-4895.
doi: 10.1080/01431160802653724
[7] Canty M J, Nielsen A A. Linear and kernel methods for multivariate change detection[J]. Computers & Geosciences, 2012, 38(1):107-114.
doi: 10.1016/j.cageo.2011.05.012
[8] Foca G, Ferrari C, Ulrici A, et al. The potential of spectral and hyperspectral-imaging techniques for bacterial detection in food:A case study on lactic acid bacteria[J]. Talanta, 2016, 2016(153):111-119.
[9] Li B, Hou B, Zhang D, et al. Pears characteristics (soluble solids content and firmness prediction,varieties) testing methods based on visible-near infrared hyperspectral imaging[J]. Optik International Journal for Light and Electron Optics, 2016, 127(5):2624-2630.
doi: 10.1016/j.ijleo.2015.11.193
[10] 韩彦岭, 崔鹏霞, 杨树瑚, 等. 基于残差网络特征融合的高光谱图像分类[J]. 国土资源遥感, 2021, 33(2):11-19.doi: 10.6046/gtzyyg.2020209.
doi: 10.6046/gtzyyg.2020209
Han Y L, Gui P X, Yang S H, et al. Classification of hyperspectral image based on feature fusion of residual network[J]. Remote Sensing for Land and Resources, 2021, 33(2):11-19.doi: 10.6046/gtzyyg.2020209.
doi: 10.6046/gtzyyg.2020209
[11] Chang Y, Wang Y, Liu T, et al. Fault diagnosis of a mine hoist using PCA and SVM techniques[J]. Journal of China University of Mining and Technology, 2008, 18(3):327-331.
doi: 10.1016/S1006-1266(08)60069-3
[12] Li C, Liu L, Lei Y, et al. Clustering for HSI hyperspectral image with weighted PCA and ICA[J]. Journal of Intelligent & Fuzzy Systems:Applications in Engineering and Technology, 2017, 32(5):3729-3737.
[13] 臧卓, 林辉, 杨敏华. 利用PCA算法进行乔木树种高光谱数据降维与分类[J]. 测绘科学, 2014, 188(2):146-149.
Zang Z, Lin H, Yang M H. Dimension reduction and classification of hyperspectral data of tree species using PCA algorithm[J]. Science of Surveying and Mapping, 2014, 188(2):146-149.
[14] 黄鸿, 曲焕鹏. 基于半监督稀疏鉴别嵌入的高光谱遥感影像分类[J]. 光学精密工程, 2014, 22(2):434-442.
Huang H, Qu H P. Hyperspectral remote sensing image classification based on SSDE[J]. Optics and Precision Engineering, 2014, 22(2):434-442.
doi: 10.3788/OPE.20142202.0434
[15] Lee C, Youn S, Jeong T, et al. Hybrid compression of hyperspectral images based on PCA with pre-encoding discriminant information[J]. IEEE Geoscience and Remote Sensing Letters, 2015, 12(7):1491-1495.
doi: 10.1109/LGRS.2015.2409897
[16] 袁林, 胡少兴, 张爱武, 等. 基于深度学习的高光谱图像分类方法[J]. 人工智能与机器人研究, 2017, 6(1):31-39.
Yuan L, Hu S X, Zhang A W, et al. A classification method for hyperspectral imagery based on deep learning[J]. Artificial Intelligence and Robotics Research, 2017, 6(1):31-39.
doi: 10.12677/AIRR.2017.61005
[17] 臧卓, 林辉, 杨敏华. ICA与PCA在高光谱数据降维分类中的对比研究[J]. 中南林业科技大学学报, 2011, 31(11):18-22.
Zang Z, Lin H, Yang M H. Comparative study on descending dimension classification of hyperspectral data between ICA algorithm and PCA algorithm[J]. Journal of Central South University of Forestry & Technology, 2011, 31(11):18-22.
[18] 叶珍, 何明一. PCA与移动窗小波变换的高光谱决策融合分类[J]. 中国图象图形学报, 2015, 20(1):132-139.
Ye Z, He M Y. PCA and windowed wavelet transform for hyperspectral decision fusion classification[J]. Journal of Image and Graphics, 2015, 20(1):132-139.
[19] Abdolmaleki M, Fathianpour N, Tabaei M. Evaluating the performance of the wavelet transform in extracting spectral alteration features from hyperspectral images[J]. International Journal of Remote Sensing, 2018, 2018:1-19.
[20] Mather P, Koch M. Computer processing of remotely-sensed images[M]. John Wiley & Sons,Ltd, 2011,265.
[21] Rodarmel C, Shan J. Principal component analysis for hyperspectral image classification[J]. Surveying and Land Information Systems, 2002, 62(2):115-122.
[22] Ibarrola-Ulzurrun E, Marcello J, Gonzalo-Martin C. Assessment of component selection strategies in hyperspectral imagery[J]. Entropy, 2017, 19:1-17.
doi: 10.3390/e19010001
[23] Stephan K, Hibbitts C, Hoffmann H, et al. Reduction of instrument-dependent noise in hyperspectral image data using the principal component analysis:Applications to Galileo NIMS data[J]. Planetary and Space Science, 2008, 56(3-4):406-419.
doi: 10.1016/j.pss.2007.11.021
[24] Zheng W, Lai J, Yuen P. GA-Fisher:A new LDA-based face recognition algorithm with selection of principal components[J]. IEEE Transactions on Systems,Man and Cybernetics,Part B (Cybernetics), 2005, 35(5):1065-1078.
doi: 10.1109/TSMCB.2005.850175
[25] Zhang X, Li R, Jiao L. Feature extraction combining PCA and immune clonal selection for hyperspectral remote sensing image classification[C]. International Conference on Artificial Intelligence & Computational Intelligence,IEEE, 2009.
[26] Tobler W. Cellular geography[M]. Springer Netherlands, 1979,379-386.
[27] 戴平生, 陈建宝. 空间统计学研究应用综述[C]// 烟台: 国际应用统计学术研讨会, 2008.
Dai P S, Chen J B. Review of spatial statistics research applications[C]// Yantai: International Symposium on Applied Statistics, 2008.
[28] Liu Q, Sun J, Chen Y, et al. Spatial variability of soil heavy metals at different sampling scales[J]. Chinese Journal of Soil Science, 2009, 40(6):1406-1410.
[29] 刘昌振, 舒红, 张志, 等. 基于多尺度分割的高分遥感图像变异函数纹理提取和分类[J]. 国土资源遥感, 2015, 27(4):47-53.doi: 10.6046/gtzyyg.2015.04.08.
doi: 10.6046/gtzyyg.2015.04.08
Liu C Z, Shu H, Zhang Z, et a1. Variogram texture extraction and classification of high resolution remote sensing images based on multi-resolution segmentation[J]. Remote Sensing for Land and Resources, 2015, 27(4):47-53.doi: 10.6046/gtzyyg.2015.04.08.
doi: 10.6046/gtzyyg.2015.04.08
[30] 张亮. 基于PCA和SVM的高光谱遥感图像分类研究[J]. 光学技术, 2008, 34(s1):184-187.
Zhang L. Study on the hyperspectral remote sensed image classify based on PCA and SVM[J]. Optical Technique, 2008, 34(s1):184-187.
[31] 王华, 李卫卫, 李志刚, 等. 基于多尺度超像素的高光谱图像分类研究[J]. 自然资源遥感, 2021, 33(3):63-71.doi: 10.6046/zrzyyg.2020344.
doi: 10.6046/zrzyyg.2020344
Wang H, Li W W, Li Z G, et al. Hyperspectral image classification based on multiscale superpixels[J]. Remote Sensing for Natural Resources 2021, 33(3):63-71.doi: 10.6046/zrzyyg.2020344.
doi: 10.6046/zrzyyg.2020344
[32] Goovaerts P. Geostatistics for natural resources evaluation[M]. NewYork City: Oxford University Press, 1997,266.
[33] Cressie N. Fitting variogram models by weighted least squares[J]. Mathematical Geology, 1985, 17(5):563-586.
[34] 矫希国, 刘超. 变差函数的参数模拟[J]. 物探化探计算技术, 1996, 18(2):157-161.
Jiao X G, Liu C. Estimation of variation parameter[J]. Computing Techniques for Geophysical and Geochemical Exploration, 1996, 18(2):157-161.
[1] 陈昊宇, 项磊, 高贺, 牟金燚, 索晓晶, 滑博伟. 基于分数阶微分的土壤全氮高光谱反演[J]. 自然资源遥感, 2023, 35(3): 170-178.
[2] 王嘉芃, 徐建国, 沈家晓, 张登荣. 德兴铜矿矿山重金属污染修复效果高光谱遥感评价[J]. 自然资源遥感, 2023, 35(3): 284-291.
[3] 梁锦涛, 陈超, 张自力, 刘志松. 一种融合指数与主成分分量的随机森林遥感图像分类方法[J]. 自然资源遥感, 2023, 35(3): 35-42.
[4] 李志新, 王梦飞, 贾伟洁, 纪松, 王宇飞. 一种结合阴影信息的建筑物层数识别方法[J]. 自然资源遥感, 2023, 35(3): 97-106.
[5] 郑宗生, 刘海霞, 王振华, 卢鹏, 沈绪坤, 唐鹏飞. 改进3D-CNN的高光谱图像地物分类方法[J]. 自然资源遥感, 2023, 35(2): 105-111.
[6] 张国建, 刘胜震, 孙英君, 俞凯杰, 刘丽娜. 基于弱监督鲁棒性自编码的高光谱异常检测[J]. 自然资源遥感, 2023, 35(2): 167-175.
[7] 赵连杰, 吴孟泉, 郑龙啸, 栾绍鹏, 赵贤峰, 薛明月, 刘佳燕, 刘晨曦. 胶东半岛北部海岸线时空变迁及驱动分析[J]. 自然资源遥感, 2022, 34(4): 87-96.
[8] 孔卓, 杨海涛, 郑逢杰, 李扬, 齐济, 朱沁雨, 杨忠霖. 高光谱遥感图像大气校正研究进展[J]. 自然资源遥感, 2022, 34(4): 1-10.
[9] 张鹏强, 高奎亮, 刘冰, 谭熊. 联合空谱信息的高光谱影像深度Transformer网络分类[J]. 自然资源遥感, 2022, 34(3): 27-32.
[10] 孙肖, 徐林林, 王晓阳, 田野, 王伟, 张中跃. 基于优化K-P-Means解混方法的高光谱图像矿物识别[J]. 自然资源遥感, 2022, 34(3): 43-49.
[11] 王春霞, 张俊, 李屹旭, Phoumilay. 复杂环境下GF-2影像水体指数的构建及验证[J]. 自然资源遥感, 2022, 34(3): 50-58.
[12] 邓静雯, 田义超, 张强, 陶进, 张亚丽, 黄升光. 机载LiDAR在红树林林分平均高估算中的应用[J]. 自然资源遥感, 2022, 34(3): 129-137.
[13] 朱琦, 郭华东, 张露, 梁栋, 刘栩婷, 万祥星. 基于多时相Landsat8影像的海南岛热带天然林类型遥感分类[J]. 自然资源遥感, 2022, 34(2): 215-223.
[14] 晏红波, 韦晚秋, 卢献健, 杨志高, 黎振宝. 基于高光谱特征的土壤含水量遥感反演方法综述[J]. 自然资源遥感, 2022, 34(2): 1-9.
[15] 王茜, 任广利. 高光谱遥感异常信息在阿尔金索拉克地区铜金矿找矿工作中的应用[J]. 自然资源遥感, 2022, 34(1): 277-285.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
京ICP备05055290号-2
版权所有 © 2015 《自然资源遥感》编辑部
地址:北京学院路31号中国国土资源航空物探遥感中心 邮编:100083
电话:010-62060291/62060292 E-mail:zrzyyg@163.com
本系统由北京玛格泰克科技发展有限公司设计开发