基于移动窗口Fourier变换的高分辨率遥感影像森林分类研究
|
摘要
以2008年4月获取的浙江省杭州市临安区东部区域分辨率为0.5 m的WorldView-1全色波段影像为数据源,在移动窗口基础上进行二维Fourier变换,构建纹理特征向量,采用不同的分类方法对森林进行分类,以寻找合适的移动窗口尺寸和分类方法。移动窗口按奇数从3×3增大到43×43,共21个不同尺寸的正方形窗口,每个边长窗口产生的纹理特征均采用Fisher判别法、随机森林、支持向量机、夹角余弦和相关系数进行分类,统计分类精度。以森林分类精度为依据,5种分类方法对应的最佳窗口依次为41×41,41×41,23×23,39×39和39×39;在最佳窗口下,5种分类方法区分森林与非森林的精度均在95%以上,总分类精度大小顺序为:Fisher判别法>随机森林>支持向量机>相关系数>夹角余弦,其中Fisher判别法总精度为99.81%,Kappa系数为0.996 3。在提取森林的基础上,进一步对森林树种(组)进行分类,总精度大小顺序为:Fisher判别法>随机森林>支持向量机>相关系数>夹角余弦,其中Fisher判别法总精度为84.86%,Kappa系数为0.8149。研究结果表明,最佳窗口下Fisher判别法的分类性能优于其他4种分类方法。
WorldView-1 panchromatic band data of 0.5 m spatial resolution image of east Lin'an district of Hangzhou, Zhejiang province in April 2008 was used as data source, and two-dimensional Fourier transform based on moving windows was carried out to produce a texture feature vector, and different classification methods were used to classify forests based on feature vectors to find an appropriate moving window size. A total of 21 square windows with odd side lengths from 3×3 to 43×43 were tested. Texture features generated by each side were classified by Fisher discriminant, random forest(RF), support vector machine(SVM), included angle cosine(IAC) and correlation coefficient(CC), and classification accuracy was computed. Based on the forest classification accuracy, the optimal windows corresponding to the five classification methods were 41×41, 41×41, 23×23, 39×39, and 39×39. Under the optimal window, five classification methods had an accuracy of 95% to distinguish forests from non-forests, and the order of the total classification accuracy was as follows: Fisher discriminant> RF > SVM > CC > IAC, the Fisher's discriminant method had a total accuracy and a Kappa coefficient of 99.81% and 0.996 3. Forest tree species were further classified, the total accuracy was Fisher discriminant> RF > SVM > CC> IAC, Fisher discriminant method had a total accuracy and Kappa coefficient of 84.86% and 0.814 9. The results showed that under the optimal window, Fisher discriminant method was the best classification method.
WorldView-1 panchromatic band data of 0.5 m spatial resolution image of east Lin'an district of Hangzhou, Zhejiang province in April 2008 was used as data source, and two-dimensional Fourier transform based on moving windows was carried out to produce a texture feature vector, and different classification methods were used to classify forests based on feature vectors to find an appropriate moving window size. A total of 21 square windows with odd side lengths from 3×3 to 43×43 were tested. Texture features generated by each side were classified by Fisher discriminant, random forest(RF), support vector machine(SVM), included angle cosine(IAC) and correlation coefficient(CC), and classification accuracy was computed. Based on the forest classification accuracy, the optimal windows corresponding to the five classification methods were 41×41, 41×41, 23×23, 39×39, and 39×39. Under the optimal window, five classification methods had an accuracy of 95% to distinguish forests from non-forests, and the order of the total classification accuracy was as follows: Fisher discriminant> RF > SVM > CC > IAC, the Fisher's discriminant method had a total accuracy and a Kappa coefficient of 99.81% and 0.996 3. Forest tree species were further classified, the total accuracy was Fisher discriminant> RF > SVM > CC> IAC, Fisher discriminant method had a total accuracy and Kappa coefficient of 84.86% and 0.814 9. The results showed that under the optimal window, Fisher discriminant method was the best classification method.
引文
[1]KIEMA J B K.Texture analysis and data fusion in the extraction of topographic objects from satellite imagery[J].Int J Remot Sens,2002,23(4):767-776.
[2]COOLEY J,LEWIS P,WELCH P.Application of the fast Fourier transform to computation of Fourier integrals,Fourier series,and convolution integrals[J].IEEE Trans Audio Electroacoust,1967,15(2):79-84.
[3]COOLEY J W,LEWIS P A W,WELCH P D.Historical notes on the fast Fourier transform[J].Proc IEEE Audio Electroacoust Trans,1967,55(10):1675-1677.
[4]ZHOU F,FENG J F,SHI Q Y.Texture feature based on local Fourier transform[C].Int Conf Image Proc,2001.Proceedings IEEE,2001,2:610-613.
[5]KAZEMI F M,SAMADI S,POORREZA H R,et al.Vehicle Recognition Based on Fourier,avelet and Curvelet Transforms-a Comparative Study[J].Int J Comput Sci Network Secur,2007,7:939-940.
[6]LI W X,ZHANG D,UNIVERSITY B,et al.Palmprint Recognition Based on Fourier Transform[J].J Software,2002,13:45-51.
[7]TSAI D M,HUANG T Y.Automated surface inspection for statistical textures[J].Image Vision Comput,2003,21(4):307-323.
[8]朱小燕,王松.傅立叶变换在粘连文字图像切分中的应用[J].计算机学报,1999,22(12):1246-1252.
[9]陈竹修.基于傅里叶变换的形状上下文描述方法[J].计算机应用与软件,2007,24(6):140-144.
[10]徐贵力,毛罕平.利用傅里叶变换提取图像纹理特征新方法[J].光电工程,2004,31(11):55-58.
[11]GAO W,HUYEN N T,LOI H S,et al.Real-time 2D parallel windowed Fourier transform for fringe pattern analysis using Graphics Processing Unit[J].Optics Express,2009,17(25):23147-23152.
[12]QIAN K.Two-dimensional windowed Fourier transform for fringe pattern analysis:Principles,applications and implementations[J].Optics Lasers Engin,2007,45(2):304-317.
[13]KEMAO Q.Windowed Fourier transform for fringe pattern analysis[J].Appl Optics,2008,43(13):2695-2702.
[14]HUANG L,QIAN K,PAN B,et al.Comparison of Fourier transform,windowed Fourier transform,and wavelet transform methods for phase extraction from a single fringe pattern in fringe projection profilometry[J].Optics Lasers Engin,2010,48(2):141-148.
[15]肖鹏峰,冯学智,等.高分辨率遥感图像分割与信息提取[M].北京:科学出版社,2012.
[16]杜靖媛,葛宏立,路伟,等.基于Fisher判别的层次分类法的森林遥感影像分类[J].西南林业大学学报,2017,37(4)175-182.
[17]方匡南,吴见彬,朱建平,等.随机森林方法研究综述[J].统计与信息论坛,2011,26(3):32-38.
[18]BREIMAN L.Random Forests[J].Machine Learn,2001,45(1):5-32.
[19]王书玉,张羽威,于振华.基于随机森林的洪河湿地遥感影像分类研究[J].测绘与空间地理信息,2014,4:83-85.
[20]田盛丰,黄厚宽.基于支持向量机的数据库学习算法[J].计算机研究与发展,2000,37(1):17-22.
[21]刘向东,陈兆乾.一种快速支持向量机分类算法的研究[J].计算机研究与发展,2004,41(8):1327-1332.
[22]黄衍,查伟雄.随机森林与支持向量机分类性能比较[J].软件,2012,33(6):107-110.
[23]卫俊霞,相里斌,高晓惠,等.基于K-均值聚类与夹角余弦法的多光谱分类算法[J].光谱学与光谱分析,2011,31(5):1357-1360.
[24]褚忠信,赵骞,黄海挺.基于相似系数的1种算法在遥感图像分类中的应用[J].中国海洋大学学报:自然科学版,2003,33(5):791-794.
[25]骆玉霞,陈焕伟.角度分类与距离分类器比较研究--以盐渍土分类为例[J].国土资源遥感,2002,14(2):46-48.
[26]徐维超.相关系数研究综述[J].广东工业大学学报,2012,29(3):12-17.
[27]GIBBONS J D,CHAKRABORTI S.Nonparametric Statistical Inference"3rd[J].J R Stat Soc,1986,149(3):93-113.
[28]FISHER R.On the“Probable Error”of a Coefficient of Correlation Deduced from a Small Sample[J].Metron,1921,4:3-32.
[29]FIELLER E C,PEARSON E S.Tests for Rank Correlation Coefficients:II[J].Biometrika,1961,48(1/2):29-40.
[30]张俊,于庆国,朱晓东,等.面向对象的高分辨率影像特征选择研究[J].测绘科学,2011,36(3):141-143.
[2]COOLEY J,LEWIS P,WELCH P.Application of the fast Fourier transform to computation of Fourier integrals,Fourier series,and convolution integrals[J].IEEE Trans Audio Electroacoust,1967,15(2):79-84.
[3]COOLEY J W,LEWIS P A W,WELCH P D.Historical notes on the fast Fourier transform[J].Proc IEEE Audio Electroacoust Trans,1967,55(10):1675-1677.
[4]ZHOU F,FENG J F,SHI Q Y.Texture feature based on local Fourier transform[C].Int Conf Image Proc,2001.Proceedings IEEE,2001,2:610-613.
[5]KAZEMI F M,SAMADI S,POORREZA H R,et al.Vehicle Recognition Based on Fourier,avelet and Curvelet Transforms-a Comparative Study[J].Int J Comput Sci Network Secur,2007,7:939-940.
[6]LI W X,ZHANG D,UNIVERSITY B,et al.Palmprint Recognition Based on Fourier Transform[J].J Software,2002,13:45-51.
[7]TSAI D M,HUANG T Y.Automated surface inspection for statistical textures[J].Image Vision Comput,2003,21(4):307-323.
[8]朱小燕,王松.傅立叶变换在粘连文字图像切分中的应用[J].计算机学报,1999,22(12):1246-1252.
[9]陈竹修.基于傅里叶变换的形状上下文描述方法[J].计算机应用与软件,2007,24(6):140-144.
[10]徐贵力,毛罕平.利用傅里叶变换提取图像纹理特征新方法[J].光电工程,2004,31(11):55-58.
[11]GAO W,HUYEN N T,LOI H S,et al.Real-time 2D parallel windowed Fourier transform for fringe pattern analysis using Graphics Processing Unit[J].Optics Express,2009,17(25):23147-23152.
[12]QIAN K.Two-dimensional windowed Fourier transform for fringe pattern analysis:Principles,applications and implementations[J].Optics Lasers Engin,2007,45(2):304-317.
[13]KEMAO Q.Windowed Fourier transform for fringe pattern analysis[J].Appl Optics,2008,43(13):2695-2702.
[14]HUANG L,QIAN K,PAN B,et al.Comparison of Fourier transform,windowed Fourier transform,and wavelet transform methods for phase extraction from a single fringe pattern in fringe projection profilometry[J].Optics Lasers Engin,2010,48(2):141-148.
[15]肖鹏峰,冯学智,等.高分辨率遥感图像分割与信息提取[M].北京:科学出版社,2012.
[16]杜靖媛,葛宏立,路伟,等.基于Fisher判别的层次分类法的森林遥感影像分类[J].西南林业大学学报,2017,37(4)175-182.
[17]方匡南,吴见彬,朱建平,等.随机森林方法研究综述[J].统计与信息论坛,2011,26(3):32-38.
[18]BREIMAN L.Random Forests[J].Machine Learn,2001,45(1):5-32.
[19]王书玉,张羽威,于振华.基于随机森林的洪河湿地遥感影像分类研究[J].测绘与空间地理信息,2014,4:83-85.
[20]田盛丰,黄厚宽.基于支持向量机的数据库学习算法[J].计算机研究与发展,2000,37(1):17-22.
[21]刘向东,陈兆乾.一种快速支持向量机分类算法的研究[J].计算机研究与发展,2004,41(8):1327-1332.
[22]黄衍,查伟雄.随机森林与支持向量机分类性能比较[J].软件,2012,33(6):107-110.
[23]卫俊霞,相里斌,高晓惠,等.基于K-均值聚类与夹角余弦法的多光谱分类算法[J].光谱学与光谱分析,2011,31(5):1357-1360.
[24]褚忠信,赵骞,黄海挺.基于相似系数的1种算法在遥感图像分类中的应用[J].中国海洋大学学报:自然科学版,2003,33(5):791-794.
[25]骆玉霞,陈焕伟.角度分类与距离分类器比较研究--以盐渍土分类为例[J].国土资源遥感,2002,14(2):46-48.
[26]徐维超.相关系数研究综述[J].广东工业大学学报,2012,29(3):12-17.
[27]GIBBONS J D,CHAKRABORTI S.Nonparametric Statistical Inference"3rd[J].J R Stat Soc,1986,149(3):93-113.
[28]FISHER R.On the“Probable Error”of a Coefficient of Correlation Deduced from a Small Sample[J].Metron,1921,4:3-32.
[29]FIELLER E C,PEARSON E S.Tests for Rank Correlation Coefficients:II[J].Biometrika,1961,48(1/2):29-40.
[30]张俊,于庆国,朱晓东,等.面向对象的高分辨率影像特征选择研究[J].测绘科学,2011,36(3):141-143.