基于支持向量机的Landsat-8影像森林类型识别研究
|
摘要
以吉林省汪清林业局天然林区为研究区,利用Landsat-8 OLI_TIRS多光谱遥感影像,结合森林资源野外调查数据,提取森林类型纹理、光谱特征参数,作为支持向量机的输入量,利用K-折交叉验证法确定最优核函数,识别森林类型,确定最优分类结果,评价分类精度,并与仅利用波段光谱特征的SVM分类结果进行精度对比。结果表明:利用纹理和光谱特征进行分类,构造SVM进行森林识别是可行的。惩罚系数C=100.0、核函数半径σ=1.000时的径向基核函数构造的支持向量机分类精度最好,总体分类精度可达89.58%,Kappa系数为0.87,单一分类精度中,阔叶林>针叶林>针阔混交林。只利用光谱特征的分类结果精度为81.26%,结合光谱和纹理特征的规律,能够提高分类精度。
In Wangqing natural forest area of Jilin Province, using Landsat-8 OLI_TIRS data,combined with the actual data of forest resources, took texture and spectral characteristic parameters as input element of support vector machine, chose kernel function through k-fold cross-validation, identified forest types and determined the optimal classification, and compared with spectral characteristic classification accuracy. As the results showed: it is feasiblethat forest classification accuracy with texture and spectral characteristic based on SVM. Penalty coefficient C=100.0, kernel radius σ=1.000,radial basis kernel function of SVM classification accuracy was the highest, overall classification accuracy was 89.58%, Kappa coefficient was 0.87, classification accuracy of broad-leaved forest was better than coniferous and coniferous forest,classification accuracy of the coniferous forest was the lowest. The accuracy of classification which only use the spectral characteristics was 81.26%. Forest classification accuracy with texture and spectral characteristic was better.
In Wangqing natural forest area of Jilin Province, using Landsat-8 OLI_TIRS data,combined with the actual data of forest resources, took texture and spectral characteristic parameters as input element of support vector machine, chose kernel function through k-fold cross-validation, identified forest types and determined the optimal classification, and compared with spectral characteristic classification accuracy. As the results showed: it is feasiblethat forest classification accuracy with texture and spectral characteristic based on SVM. Penalty coefficient C=100.0, kernel radius σ=1.000,radial basis kernel function of SVM classification accuracy was the highest, overall classification accuracy was 89.58%, Kappa coefficient was 0.87, classification accuracy of broad-leaved forest was better than coniferous and coniferous forest,classification accuracy of the coniferous forest was the lowest. The accuracy of classification which only use the spectral characteristics was 81.26%. Forest classification accuracy with texture and spectral characteristic was better.
引文
[1]Franklin S E.Remotesensing for sustain able forest management[M].CRC Press,2001.
[2]Nazeer M,Nichol J E,Yung Y K.Evaluation of atmospheric correction models and Landsat surface reflectance product in an urban coastal environment[J].International Journal of Remote Sensing,2014,35(16):6271-6291.
[3]Lu D,Weng Q.A survey of image classification methods and techniques for improving classification performance[J].International journal of Remote sensing,2007,28(5):823-870.
[4]Liaw A,Wiener M.Classification and regression by random Forest[J].R news,2002,2(3):18-22.
[5]张淑芬,邢艳秋,艾合买提江·阿不都艾尼,等.基于TM遥感影像的森林类型分类方法比较[J].森林工程,2014,30(1):18-21.
[6]Chang C,Xiaoyang Y,Yamei B.The Image Multi Feature Retrieval based on SVM Semantic Classification[J].International Journal of Hybrid Information Technology,2016,9(3):291-300.
[7]Tarabalka Y,Fauvel M,Chanussot J,et al.SVM-and MRF-based method for accurate classification of hyperspectral images[J].Geoscience and Remote Sensing Letters,IEEE,2010,7(4):736-740.
[8]Roy D P,Wulder M A,Loveland T R,et al.Landsat-8:Science and product vision for terrestrial global change research[J].Remote Sensing of Environment,2014(145):154-172.
[9]GB/T 14721-2010,林业资源分类与代码[S].森林类型,2010.
[10]陈玲,郝文乾,高德亮.光学影像纹理信息在林业领域的最新应用进展[J].北京林业大学学报,2015,37(3):1-12.
[11]Coburn C A,Roberts A C B.A multiscale texture analysis procedure for improved forest stand classification[J].International journal of remote sensing,2004,25(20):4287-4308.
[12]Ponti M,NazaréT S,ThuméG S.Image quantization as a dimensionality reduction procedure in color and texture feature extraction[J].Neurocomputing,2016(173):385-396.
[13]Poona N K,van Niekerk A,Nadel R L,et al.Random Forest(RF)Wrappers for Waveband Selection and Classification of Hyperspectral Data[J].Applied spectroscopy,2016,70(2):322-333.
[14]Zhao W,Du S.Learning multiscale and deep representations for classifying remotely sensed imagery[J].ISPRS Journal of Photogrammetry and Remote Sensing,2016(113):155-165.
[15]严恩萍,林辉,莫登奎,等.基于光谱特征的森林类型识别研究[J].中南林业科技大学学报,2011,31(11):23-29.
[16]Yang X.Parameterizing support vector machines for land cover classification[J].Photogrammetric Engineering&Remote Sensing,2011,77(1):27-37.
[17]刘美爽,邢艳秋,李立存,等.基于星载激光雷达数据和支持向量分类机方法的森林类型识别[J].东北林业大学学报,2014,42(2):124-128.
[18]Sun N,Chen Q X,Luo J C,et al.Coupling GA with SVMfor feature selection in high-resolution remote sensing target recognition[J].Journal of Remote Sensing,2010,14(5):928-943.
[19]Aghabalaei A,Maghsoudi Y,Ebadi H.Forest classification using extracted Pol SAR features from Compact Polarimetrydata[J].Advances in Space Research,2016,57(9):1939-1950.
[20]Huang X,Zhang L.An SVM ensemble approach combining spectral,structural,and semantic features for the classification of high-resolution remotely sensed imagery[J].Geoscience and Remote Sensing,IEEE Transactions on,2013,51(1):257-272.
[2]Nazeer M,Nichol J E,Yung Y K.Evaluation of atmospheric correction models and Landsat surface reflectance product in an urban coastal environment[J].International Journal of Remote Sensing,2014,35(16):6271-6291.
[3]Lu D,Weng Q.A survey of image classification methods and techniques for improving classification performance[J].International journal of Remote sensing,2007,28(5):823-870.
[4]Liaw A,Wiener M.Classification and regression by random Forest[J].R news,2002,2(3):18-22.
[5]张淑芬,邢艳秋,艾合买提江·阿不都艾尼,等.基于TM遥感影像的森林类型分类方法比较[J].森林工程,2014,30(1):18-21.
[6]Chang C,Xiaoyang Y,Yamei B.The Image Multi Feature Retrieval based on SVM Semantic Classification[J].International Journal of Hybrid Information Technology,2016,9(3):291-300.
[7]Tarabalka Y,Fauvel M,Chanussot J,et al.SVM-and MRF-based method for accurate classification of hyperspectral images[J].Geoscience and Remote Sensing Letters,IEEE,2010,7(4):736-740.
[8]Roy D P,Wulder M A,Loveland T R,et al.Landsat-8:Science and product vision for terrestrial global change research[J].Remote Sensing of Environment,2014(145):154-172.
[9]GB/T 14721-2010,林业资源分类与代码[S].森林类型,2010.
[10]陈玲,郝文乾,高德亮.光学影像纹理信息在林业领域的最新应用进展[J].北京林业大学学报,2015,37(3):1-12.
[11]Coburn C A,Roberts A C B.A multiscale texture analysis procedure for improved forest stand classification[J].International journal of remote sensing,2004,25(20):4287-4308.
[12]Ponti M,NazaréT S,ThuméG S.Image quantization as a dimensionality reduction procedure in color and texture feature extraction[J].Neurocomputing,2016(173):385-396.
[13]Poona N K,van Niekerk A,Nadel R L,et al.Random Forest(RF)Wrappers for Waveband Selection and Classification of Hyperspectral Data[J].Applied spectroscopy,2016,70(2):322-333.
[14]Zhao W,Du S.Learning multiscale and deep representations for classifying remotely sensed imagery[J].ISPRS Journal of Photogrammetry and Remote Sensing,2016(113):155-165.
[15]严恩萍,林辉,莫登奎,等.基于光谱特征的森林类型识别研究[J].中南林业科技大学学报,2011,31(11):23-29.
[16]Yang X.Parameterizing support vector machines for land cover classification[J].Photogrammetric Engineering&Remote Sensing,2011,77(1):27-37.
[17]刘美爽,邢艳秋,李立存,等.基于星载激光雷达数据和支持向量分类机方法的森林类型识别[J].东北林业大学学报,2014,42(2):124-128.
[18]Sun N,Chen Q X,Luo J C,et al.Coupling GA with SVMfor feature selection in high-resolution remote sensing target recognition[J].Journal of Remote Sensing,2010,14(5):928-943.
[19]Aghabalaei A,Maghsoudi Y,Ebadi H.Forest classification using extracted Pol SAR features from Compact Polarimetrydata[J].Advances in Space Research,2016,57(9):1939-1950.
[20]Huang X,Zhang L.An SVM ensemble approach combining spectral,structural,and semantic features for the classification of high-resolution remotely sensed imagery[J].Geoscience and Remote Sensing,IEEE Transactions on,2013,51(1):257-272.