CBERS CCD数据土地利用/覆被分类研究
|
摘要
目前,卫星遥感应用的水平常滞后于空间遥感技术的发展,主要表现在:①卫星发回的遥感数据未得到充分利用,据统计,目前卫星发回的遥感数据得到应用的不足其总量的10%;②对遥感信息认识的不足和对遥感专题信息分析水平的滞后,造成了遥感信息资源的巨大浪费及其应用价值的降低。这主要在于缺少完善的遥感图像分析处理的方法和模型,难以从遥感图像中直接获得大量高精度的空间信息。因此挖掘卫星遥感信息的应用潜力,特别是国产资源卫星数据,提高遥感图像分析识别的精度,是目前遥感应用的迫切要求。
本文选取CBERS多光谱数据,以江苏省宜兴地区作为实验区,首先进行了CBERS-02星多光谱数据的图像预处理,然后利用ERDAS IMAGINE和Definiens7.0等软件,采用非监督分类、监督分类、面向对象的方法进行了土地利用/覆被分类,并将影像分类结果进行了精度评价。
结果表明:与传统监督和非监督方法相比,面向对象的分类方法在利用光谱信息的同时,还考虑了影像对象的空间信息,使分类结果有了明显改进。具体来讲,在一级地类的提取中,面向对象的最邻近分类器的分类精度最低,监督分类的精度最高;在二级地类的提取中,面向对象的特征阈值法得到的分类结果较好,能够轻松提取传统分类方法难于提取的地类。这是因为最邻近分类器的特征空间只设置了与光谱信息有关的特征,面向对象的特征阈值法考虑了影像对象的空间信息,如果在此基础上利用Definiens7.0软件提供的人机交互工具进行目视修正还可以进一步提高分类精度。
面向对象分类方法可以生成带有属性列表的专题栅格层,也可以是带有属性信息的矢量多边形,可用于进行GIS的空间分析和决策支持,这为GIS和RS集成提供了思路。
Currently, the application of satellite remote sensing have been leg behind the development of space space remote sensing, especially in the following aspect: 1) remote sensing data sent by satellite are not fully used, according to statistics, only 10% of the total data were used; 2) insufficient recognize of the remote sensing information and leg of Information Analyst of Remote Sensing Special Subject. These are mainly due to lack incomplete method and model of remote sensing image analysis, hard to extract high accuracy space information from satellite image.
Therefore, it is pressed for sensing application to dig the application potential of satellite remote sensing information, especially to domestic satellite data. Taking Yixing City of Jiangsu Province as a test field and based on multi-spectral remote sensing image data of CBERS-02 CCD, this paper firstly made a series of image pre-processing. Then, by using ERDAS IMAGINE and Definiens7.0, the land use categories were extracted by unsupervised classification, supervised classification, integration of supervised classification and visual interpretation, object-oriented approach. Finally, this paper made a comparison of the four classification precision evaluation.
Experiment indicated that: compared with supervised and unsupervised classification, object-oriented method not only consider spectral information, but also consider space information of the image object, which have improve the classification results. More specifically, among the extraction of prime types, the lowest classification precision is Nearest Neighbor Classifier of object-oriented method, and the highest classification precision is supervised classification. Among the extraction of secondary types, methods of based on Features of object-oriented method have a relatively satisfaction result, which could easily extract some secondary types which are hard to be extracted by traditional methods. The reason is that the feature space of Nearest Neighbor Classifier only make use of spectral information in this experiment, in other words, it did not make use of space information. When it comes to the extraction of the secondary types, both the spectral information and space information are considered. Based on this result, utilized the Manual Editor tool of Definiens7.0, we still can improve the results of the classification.
Object-oriented method can generate grid layer with attribute information as well as vector polygon with attribute information, which can do help to spatial analysis and decision support, which provide a good way to the integration of GIS and RS.
本文选取CBERS多光谱数据,以江苏省宜兴地区作为实验区,首先进行了CBERS-02星多光谱数据的图像预处理,然后利用ERDAS IMAGINE和Definiens7.0等软件,采用非监督分类、监督分类、面向对象的方法进行了土地利用/覆被分类,并将影像分类结果进行了精度评价。
结果表明:与传统监督和非监督方法相比,面向对象的分类方法在利用光谱信息的同时,还考虑了影像对象的空间信息,使分类结果有了明显改进。具体来讲,在一级地类的提取中,面向对象的最邻近分类器的分类精度最低,监督分类的精度最高;在二级地类的提取中,面向对象的特征阈值法得到的分类结果较好,能够轻松提取传统分类方法难于提取的地类。这是因为最邻近分类器的特征空间只设置了与光谱信息有关的特征,面向对象的特征阈值法考虑了影像对象的空间信息,如果在此基础上利用Definiens7.0软件提供的人机交互工具进行目视修正还可以进一步提高分类精度。
面向对象分类方法可以生成带有属性列表的专题栅格层,也可以是带有属性信息的矢量多边形,可用于进行GIS的空间分析和决策支持,这为GIS和RS集成提供了思路。
Currently, the application of satellite remote sensing have been leg behind the development of space space remote sensing, especially in the following aspect: 1) remote sensing data sent by satellite are not fully used, according to statistics, only 10% of the total data were used; 2) insufficient recognize of the remote sensing information and leg of Information Analyst of Remote Sensing Special Subject. These are mainly due to lack incomplete method and model of remote sensing image analysis, hard to extract high accuracy space information from satellite image.
Therefore, it is pressed for sensing application to dig the application potential of satellite remote sensing information, especially to domestic satellite data. Taking Yixing City of Jiangsu Province as a test field and based on multi-spectral remote sensing image data of CBERS-02 CCD, this paper firstly made a series of image pre-processing. Then, by using ERDAS IMAGINE and Definiens7.0, the land use categories were extracted by unsupervised classification, supervised classification, integration of supervised classification and visual interpretation, object-oriented approach. Finally, this paper made a comparison of the four classification precision evaluation.
Experiment indicated that: compared with supervised and unsupervised classification, object-oriented method not only consider spectral information, but also consider space information of the image object, which have improve the classification results. More specifically, among the extraction of prime types, the lowest classification precision is Nearest Neighbor Classifier of object-oriented method, and the highest classification precision is supervised classification. Among the extraction of secondary types, methods of based on Features of object-oriented method have a relatively satisfaction result, which could easily extract some secondary types which are hard to be extracted by traditional methods. The reason is that the feature space of Nearest Neighbor Classifier only make use of spectral information in this experiment, in other words, it did not make use of space information. When it comes to the extraction of the secondary types, both the spectral information and space information are considered. Based on this result, utilized the Manual Editor tool of Definiens7.0, we still can improve the results of the classification.
Object-oriented method can generate grid layer with attribute information as well as vector polygon with attribute information, which can do help to spatial analysis and decision support, which provide a good way to the integration of GIS and RS.
引文
[1] Hinton J.C. Image Classification and Analysis Using Integrated GIS. Advances in Remote Sensing and GIS Aanlysis.1999,207-219.
[2] Schowengedt R.A. Remote Sensing model and methods for image processing.1997,2-11.
[3] Congalton R.G.,etal. Mapping and monitoring agricultural crops and other land cover in the lower Colorado River Basin.PE&RS,1998,64(11):1107-1113.
[4] Stoney W. A New Space Race Is On.GIS World. 1998.11(3):44-46.
[5]何国金,胡德永.卫星遥感数据的信息论理解地质科技情报.1997,16(增刊):43-47.
[6]杨冰,刘成武,黄利民.土地利用/土地覆被变化的研究方法[J].地域研究与开发,2004,Vol.23(4):11-15.
[7]陈佑启,杨鹏.国际上土地利用/土地覆被变化研究的新进展[J].经济地理,2001,Vol.21(1):95-9.
[8] AL Gore. The Digital Earth: Understanding our planet in the 21st Century at the California Science Center, Los Angeles, California, on January 31, 1998.
[9]胡如忠.中巴地球资源卫星及其应用[M].北京:遥感应用协会,2002:7～8,11～16.
[10]中国资源卫星应用中心网站[EB/OL].http://www.cresda.cn.
[11]中国科学院中国遥感卫星地面站网站[EB/OL].http://www.rsgs.ac.cn.
[12]张庆军,马世俊.中国航天2008年第4期.
[13]陈全育.总师细说中巴资源一号02B星《太空探索》2007年第11期.
[14]资源一号卫星02B在云南香格里拉县森林火灾监测中发挥重要作用。中国资源卫星应用中心简报, 2008年第11期.
[15]张庆君,马世俊.中巴地球资源卫星技术特点及技术进步。中国航天2008年第4期13-18.
[16]李四海,刘振民等.中巴卫星数据在海岸带环境监测中的应用[J].遥感技术与应用,2003,18(2):66～72.
[17]丁峰,高志海等. CBERS-1数据评价及在荒漠化监测中的应用[J].遥感技术与应用,2004,19 (5) :339～342.
[18]党福星,方洪宾等.利用CBERS-1CCD数据进行地质矿产信息提取方法研究[J].国土资源遥感,2002 , 54 (4) :63～66.
[19]韩爱惠,王庆杰等. CBERS-02星CCD数据在林业资源监测中的应用评价.国土资源遥感[J ] . 2004 (2) :61～6.
[20]陈述彭,胡如忠等. CBERS-1卫星图像在黄河三角洲可持续发展中的应用研究[J ].航天返回与遥感,2001 ,22 (3) :34～39.
[21]罗扬,朱军等.中巴资源卫星数据在贵州森林资源调查中的应用[J].南京林业大学学报(自然科学版) ,2005 ,29 (2) :91～94.
[22] Zhengjun Liu , Aixia Liu, Changyao Wang. Evolving neural net2work using real coded genetic algorithm (GA) for multi-spectral image classification. Future Generation Computer Systems ,2004 ,(20) :1119～1129.
[23]骆成凤,王长耀.用模糊ARTMAP算法对CBERS-2数据进行分类[J].武汉大学学报(信息科学版) ,2006 , 31 (4) :325～328.
[24]夏建勋,韩小明等.中巴资源卫星数据影像在新疆国土资源环境综合遥感调查项目中的应用[J].新疆地质,2003 ,21 (3) :362～363.
[25]丁峰,高志海. CBERS-1数据评价及在荒漠化监测中的应用[J].遥感技术与应用,2004 ,19 (5) :339～342.
[26]李小英,顾行发等.CBERS-02 WFI的辐射交叉定标及其对植被指数的作用[J].遥感学报,2006,12 3):211～220.
[27]赵英时.遥感应用分析原理与方法[M].北京:科学出版社,2005.第194页.
[28]杜培军.遥感原理与应用[M].徐州:中国矿业大学出版社,2006.第166页.
[29]袁金国.遥感图像数字处理[M].北京:中国环境出版社,2006.
[30]曾如珠.遥感图像分类识别的探讨[J].泉州师范学院学报(自然科学),2007(7):36-39.
[31]冯仲科,余新晓.“3S”技术及其利用[M]..北京:中国林业出版社,2001.
[32]何宇华,史良树,张荣慧,等.中巴资源卫星数据(CBERS-02)在土地调查中的应用[J].中国土地科学,2007,(2):51～57.
[33] Berberoglu S, etc. The integration of spectral and textural information using neural networks for land cover mapping in the Mediterranean. Computers & Geosciences.200, 26: 385- 396.
[34] Wang D.L, Terman. Locally excitatory globally inhibitory oscillators network: theory and application to pattern segmentation. IEEE International Conference of Neural Network,1994(2)945-950.
[35]李朝锋,曾生根,许磊.遥感图像智能处理[M].北京:电子工业出版社,2007.第6页.
[36] Vapnik V. The Nature of Statistical Learning Theory. New York: Springer-Verlag,1995.
[37] Burges CJC. A tutorial on Support Vector Machines for pattern recognition. DATA MINING AND KNOWLEDGE DISCOVERY,19812(2).
[38] Gualtieri JA, Cromp RF. Support vector machines for hyper-spectral remote sensing classification. : PROCEEDINGS OF THE SPIE.
[39] Huang C, Davis LS, Townshend JRG. INTERNATIONAL JOURNAL OF REMOTE SENSING,2002,23(4):725-749 ,Nemmour H, Chibani Y.
[40] Nemmour H, Chibani Y. Multiple support vector machines for land cover change detection: An application for mapping urban extensions. ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING.61 (2006)125-133.
[41]骆剑承,周成虎等.支撑向量机及其遥感影像空间特征提取和分类的应用研究[J].遥感学报,2002,6(1):50-55.
[42]史培军.土地利用/土地覆被变化研究的方法与实践[M].北京:科学出版社,2000.
[43]朱述龙,朱宝山,王红卫.遥感图像处理与应用[M].北京:科学出版社,2006.
[44]宜兴年鉴. 1996 [专著] /宜兴市地方志编纂委员会办公室编辑.南京:江苏人民出版社, 1997.
[45]陈百明等,土地资源学[M].北京:北京师范大学出版社2008:33～34.
[46]赵文慧,赵文吉,李小娟,等.中巴资源二号卫星影像在土地利用变化中的应用[J].首都师范大学学报(自然科学版),2007,(6):78～82.
[47]周成虎,骆剑承,杨晓梅,等.遥感影像地学理解与分析[M].北京:科学出版社, 2003 :41～53.
[48]于海洋,甘甫平,党福星.高分辨率遥感影像波段配准误差试验分析[J].国土资源遥感, 2007,(3):39～42.
[49]党安荣,王晓栋,陈晓峰等.ERDAS IMAGINE遥感图像处理方法[M].北京:清华大学出版社,2003:186～219.
[50]霍东民,严明,于冰洋.DMC遥感小卫星数据预处理方法分析[J].遥感学报,2005,9(4):480～485.
[51]潘建刚,赵文吉,宫辉力.遥感图像分类方法的研究[J].首都师范大学学报,2004,25 (3):86～91.
[52]李石华,王金亮,毕艳,等.遥感图像分类方法研究综述[J].国土资源遥感,2005,64(2):1～5.
[53]何宇华,史良树,张荣慧,等.中巴资源卫星数据(CBERS-02)在土地调查中的应用[J].中国土地科学,2007,(2):51～57.
[54] Frank Lin S E, Wilson B A. Spatial and spectral classification for remote～sensing imagery [J]. Computers and Geosciences , 1991,(17):1151～1172.
[55]Haralick RM,Shapiro LG. Image segmentation techniques [J]. Computer, Vision, Graphics and Image Processing, 1985, (29):100～132.
[56]曹雪,柯长青.基于对象级的高分辨率遥感影像分类研究.遥感信息,2006(5) :27-30.
[2] Schowengedt R.A. Remote Sensing model and methods for image processing.1997,2-11.
[3] Congalton R.G.,etal. Mapping and monitoring agricultural crops and other land cover in the lower Colorado River Basin.PE&RS,1998,64(11):1107-1113.
[4] Stoney W. A New Space Race Is On.GIS World. 1998.11(3):44-46.
[5]何国金,胡德永.卫星遥感数据的信息论理解地质科技情报.1997,16(增刊):43-47.
[6]杨冰,刘成武,黄利民.土地利用/土地覆被变化的研究方法[J].地域研究与开发,2004,Vol.23(4):11-15.
[7]陈佑启,杨鹏.国际上土地利用/土地覆被变化研究的新进展[J].经济地理,2001,Vol.21(1):95-9.
[8] AL Gore. The Digital Earth: Understanding our planet in the 21st Century at the California Science Center, Los Angeles, California, on January 31, 1998.
[9]胡如忠.中巴地球资源卫星及其应用[M].北京:遥感应用协会,2002:7～8,11～16.
[10]中国资源卫星应用中心网站[EB/OL].http://www.cresda.cn.
[11]中国科学院中国遥感卫星地面站网站[EB/OL].http://www.rsgs.ac.cn.
[12]张庆军,马世俊.中国航天2008年第4期.
[13]陈全育.总师细说中巴资源一号02B星《太空探索》2007年第11期.
[14]资源一号卫星02B在云南香格里拉县森林火灾监测中发挥重要作用。中国资源卫星应用中心简报, 2008年第11期.
[15]张庆君,马世俊.中巴地球资源卫星技术特点及技术进步。中国航天2008年第4期13-18.
[16]李四海,刘振民等.中巴卫星数据在海岸带环境监测中的应用[J].遥感技术与应用,2003,18(2):66～72.
[17]丁峰,高志海等. CBERS-1数据评价及在荒漠化监测中的应用[J].遥感技术与应用,2004,19 (5) :339～342.
[18]党福星,方洪宾等.利用CBERS-1CCD数据进行地质矿产信息提取方法研究[J].国土资源遥感,2002 , 54 (4) :63～66.
[19]韩爱惠,王庆杰等. CBERS-02星CCD数据在林业资源监测中的应用评价.国土资源遥感[J ] . 2004 (2) :61～6.
[20]陈述彭,胡如忠等. CBERS-1卫星图像在黄河三角洲可持续发展中的应用研究[J ].航天返回与遥感,2001 ,22 (3) :34～39.
[21]罗扬,朱军等.中巴资源卫星数据在贵州森林资源调查中的应用[J].南京林业大学学报(自然科学版) ,2005 ,29 (2) :91～94.
[22] Zhengjun Liu , Aixia Liu, Changyao Wang. Evolving neural net2work using real coded genetic algorithm (GA) for multi-spectral image classification. Future Generation Computer Systems ,2004 ,(20) :1119～1129.
[23]骆成凤,王长耀.用模糊ARTMAP算法对CBERS-2数据进行分类[J].武汉大学学报(信息科学版) ,2006 , 31 (4) :325～328.
[24]夏建勋,韩小明等.中巴资源卫星数据影像在新疆国土资源环境综合遥感调查项目中的应用[J].新疆地质,2003 ,21 (3) :362～363.
[25]丁峰,高志海. CBERS-1数据评价及在荒漠化监测中的应用[J].遥感技术与应用,2004 ,19 (5) :339～342.
[26]李小英,顾行发等.CBERS-02 WFI的辐射交叉定标及其对植被指数的作用[J].遥感学报,2006,12 3):211～220.
[27]赵英时.遥感应用分析原理与方法[M].北京:科学出版社,2005.第194页.
[28]杜培军.遥感原理与应用[M].徐州:中国矿业大学出版社,2006.第166页.
[29]袁金国.遥感图像数字处理[M].北京:中国环境出版社,2006.
[30]曾如珠.遥感图像分类识别的探讨[J].泉州师范学院学报(自然科学),2007(7):36-39.
[31]冯仲科,余新晓.“3S”技术及其利用[M]..北京:中国林业出版社,2001.
[32]何宇华,史良树,张荣慧,等.中巴资源卫星数据(CBERS-02)在土地调查中的应用[J].中国土地科学,2007,(2):51～57.
[33] Berberoglu S, etc. The integration of spectral and textural information using neural networks for land cover mapping in the Mediterranean. Computers & Geosciences.200, 26: 385- 396.
[34] Wang D.L, Terman. Locally excitatory globally inhibitory oscillators network: theory and application to pattern segmentation. IEEE International Conference of Neural Network,1994(2)945-950.
[35]李朝锋,曾生根,许磊.遥感图像智能处理[M].北京:电子工业出版社,2007.第6页.
[36] Vapnik V. The Nature of Statistical Learning Theory. New York: Springer-Verlag,1995.
[37] Burges CJC. A tutorial on Support Vector Machines for pattern recognition. DATA MINING AND KNOWLEDGE DISCOVERY,19812(2).
[38] Gualtieri JA, Cromp RF. Support vector machines for hyper-spectral remote sensing classification. : PROCEEDINGS OF THE SPIE.
[39] Huang C, Davis LS, Townshend JRG. INTERNATIONAL JOURNAL OF REMOTE SENSING,2002,23(4):725-749 ,Nemmour H, Chibani Y.
[40] Nemmour H, Chibani Y. Multiple support vector machines for land cover change detection: An application for mapping urban extensions. ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING.61 (2006)125-133.
[41]骆剑承,周成虎等.支撑向量机及其遥感影像空间特征提取和分类的应用研究[J].遥感学报,2002,6(1):50-55.
[42]史培军.土地利用/土地覆被变化研究的方法与实践[M].北京:科学出版社,2000.
[43]朱述龙,朱宝山,王红卫.遥感图像处理与应用[M].北京:科学出版社,2006.
[44]宜兴年鉴. 1996 [专著] /宜兴市地方志编纂委员会办公室编辑.南京:江苏人民出版社, 1997.
[45]陈百明等,土地资源学[M].北京:北京师范大学出版社2008:33～34.
[46]赵文慧,赵文吉,李小娟,等.中巴资源二号卫星影像在土地利用变化中的应用[J].首都师范大学学报(自然科学版),2007,(6):78～82.
[47]周成虎,骆剑承,杨晓梅,等.遥感影像地学理解与分析[M].北京:科学出版社, 2003 :41～53.
[48]于海洋,甘甫平,党福星.高分辨率遥感影像波段配准误差试验分析[J].国土资源遥感, 2007,(3):39～42.
[49]党安荣,王晓栋,陈晓峰等.ERDAS IMAGINE遥感图像处理方法[M].北京:清华大学出版社,2003:186～219.
[50]霍东民,严明,于冰洋.DMC遥感小卫星数据预处理方法分析[J].遥感学报,2005,9(4):480～485.
[51]潘建刚,赵文吉,宫辉力.遥感图像分类方法的研究[J].首都师范大学学报,2004,25 (3):86～91.
[52]李石华,王金亮,毕艳,等.遥感图像分类方法研究综述[J].国土资源遥感,2005,64(2):1～5.
[53]何宇华,史良树,张荣慧,等.中巴资源卫星数据(CBERS-02)在土地调查中的应用[J].中国土地科学,2007,(2):51～57.
[54] Frank Lin S E, Wilson B A. Spatial and spectral classification for remote～sensing imagery [J]. Computers and Geosciences , 1991,(17):1151～1172.
[55]Haralick RM,Shapiro LG. Image segmentation techniques [J]. Computer, Vision, Graphics and Image Processing, 1985, (29):100～132.
[56]曹雪,柯长青.基于对象级的高分辨率遥感影像分类研究.遥感信息,2006(5) :27-30.