基于WorldView-2数据的基塘系统遥感分类研究
|
摘要
分布在热带和亚热带低洼地区的基塘系统是一种典型的人工复合农业生态系统。通过水陆交互作用和能量多级利用,基塘系统创造出较高的生态价值和经济价值。掌握基塘系统土地利用的类型、面积和比例关系,有助于了解和分析基塘系统的现状及其生态模式和功能。以广东省佛山市顺德区西南部为研究区,利用高空间分辨率WorldView-2卫星影像数据,采用结合面向对象分类和基于像元的混合分类方法,提取基塘系统中的塘面和基面(基面又细分为道路、植被和裸土等);从分类特征组合与分类器两个方面,通过对比分析,筛选出基塘系统遥感影像分类的最优方法;根据最优分类结果,计算出塘基与塘面的面积比例。研究结果表明,采用面向对象分类方法分离的塘面和塘基的总体精度为82.2%;在基于像元进一步对塘基进行分类的过程中,结合8个多光谱波段光谱反射率平均值、影像亮度值和8个红边波段指数特征组合方式的随机森林分类器的分类精度最高,达到85.6%;得到的基-塘面积比例约为4.9∶5.1,植被-塘面的面积比例约为3.5∶6.5。提出的结合面向对象和基于像元的混合分类方法能够对基塘系统进行精细的遥感影像分类,能准确估计基塘系统中各地物的面积和基-塘比例等指标,并为基塘系统生态功能的调查和基塘系统的合理保护与开发提供参考和数据支持。
Dike-pond system is a typical artificial mixed agricultural ecosystem, mostly distributed in tropical and subtropical areas. Dike-pond system creates great ecological and economic values due to its water-land interactive agriculture and multi-level energy exploitation. It is essential to investigate the land use types, their areas, and proportions in dike-pond system as a base to further analyze its current situation, and ecological mechanism and functions. In this study, a study area was selected in the southwestern Shunde District in Foshan city, Guangdong, China. A multi-level image classification approach that combines object-oriented and pixel-based classification methods was proposed to extract the information of pond, dike, and land cover types on the dike, through high-resolution WorldView-2 satellite image data. By optimizing feature space and comparing Bayes, KNN, and Random Forest classifier, the classification result with the highest overall accuracy was generated and further applied to estimate the area of each kind of land cover type and the ratio between the pond and dike areas. The result showed that the overall classification accuracy of pond and dike by the first-level object-oriented classification was 82.2%. In the second-level pixel-based classification further applied only to the dike, random forest classifier with the feature space comprised of the brightness of pixels, 8 multi-spectral bands, and 8 derived indices from red-edge bands generated the classification result with the highest overall accuracy, approximately 85.6%. A further estimation indicated that the dike-pond area ratio and the vegetation-pond area ratio were approximately 4.9∶5.1 and 3.5∶6.5, respectively. The multi-level image classification approach that combines object-oriented and pixel-based classification methods in this study generated the land use maps with high-accurate dike-pond system. The area statistics and pond-dike ratios estimated from the classification results could provide data support and decision reference for the protection and development of dike-pond system.
Dike-pond system is a typical artificial mixed agricultural ecosystem, mostly distributed in tropical and subtropical areas. Dike-pond system creates great ecological and economic values due to its water-land interactive agriculture and multi-level energy exploitation. It is essential to investigate the land use types, their areas, and proportions in dike-pond system as a base to further analyze its current situation, and ecological mechanism and functions. In this study, a study area was selected in the southwestern Shunde District in Foshan city, Guangdong, China. A multi-level image classification approach that combines object-oriented and pixel-based classification methods was proposed to extract the information of pond, dike, and land cover types on the dike, through high-resolution WorldView-2 satellite image data. By optimizing feature space and comparing Bayes, KNN, and Random Forest classifier, the classification result with the highest overall accuracy was generated and further applied to estimate the area of each kind of land cover type and the ratio between the pond and dike areas. The result showed that the overall classification accuracy of pond and dike by the first-level object-oriented classification was 82.2%. In the second-level pixel-based classification further applied only to the dike, random forest classifier with the feature space comprised of the brightness of pixels, 8 multi-spectral bands, and 8 derived indices from red-edge bands generated the classification result with the highest overall accuracy, approximately 85.6%. A further estimation indicated that the dike-pond area ratio and the vegetation-pond area ratio were approximately 4.9∶5.1 and 3.5∶6.5, respectively. The multi-level image classification approach that combines object-oriented and pixel-based classification methods in this study generated the land use maps with high-accurate dike-pond system. The area statistics and pond-dike ratios estimated from the classification results could provide data support and decision reference for the protection and development of dike-pond system.
引文
[1]郭程轩,徐颂军.中国基塘系统研究进展与新视角[J].湿地科学, 2011, 99(1):75-81.
[2]钟功甫,蔡国雄.我国基(田)塘系统生态模式模式以珠江三角洲和长江三角洲为例[J].生态经济, 1987,(3):15-20.
[3]钟功甫.基塘系统的特征及其实践意义[J].地理科学, 1988, 8(1):12-17.
[4]钟功甫,王增骐,吴厚水.基塘系统的水陆相互作用[M].北京:科学出版社, 1993.
[5]聂呈荣,骆世明,章家恩,等.现代集约农业下基塘系统的退化与生态恢复[J].生态学报, 2003, 2323(9):1851-1860.
[6]丁疆华,温琰茂,舒强.基塘系统可持续发展现状、问题及对策研究[J].重庆环境科学, 2001, 2323(5):12-14.
[7]祝功武.珠三角人工地貌水文调节机制的今日思考[J].热带地理, 2012, 3232(4):378-384.
[8]袁兰,胡月明,范建彬.珠江三角洲农户整治基塘土地的意愿及影响因素研究[J].广东农业科学, 2015, 4242(7):180-186.
[9]刘青勇,王爱芹,张娜,等.基于调水调沙的黄河三角洲湿地生态修复技术[J].中国农村水利水电, 2016,(2):60-63.
[10]赵玉环.社会经济发展对珠江三角洲基塘系统的影响[J].仲恺农业技术学院学报, 2001, 1414(3):28-33.
[11]吴满强.历史时期佛山地区土地利用状况探析[D].广州:暨南大学, 2012.
[12]吴建新.明清民国顺德的基塘农业与经济转型[J].古今农业,2011,(1):96-104.
[13]杨丹.佛山市基塘景观破碎化动力机制及其模拟[D].南昌:东华理工大学, 2015.
[14]李梅,聂呈荣,龙兴.基塘系统生态环境质量评价指标体系的构建[J].农业环境科学学报, 2007, 2626(1):386-390.
[15]王晓锋,袁兴中,周李磊,等.三峡水库消落带景观基塘系统对地表径流氮磷拦截效益——以重庆开州区汉丰湖为例[J].三峡生态环境监测, 2017,(3):40-48.
[16]韩西丽,俞孔坚,李迪华,等.基塘-城市景观安全格局构建研究——以佛山市顺德区马岗片区为例[J].地域研究与开发,2008, 2727(5):107-110.
[17]林媚珍,冯荣光,纪少婷.中山市基塘农业模式演变及景观格局分析[J].广东农业科学, 2014, 4141(24):184-189.
[18]魏兴琥,王海,刁华涛.佛山市顺德区基塘农业模式的演变与发展潜力[J].佛山科学技术学院学报:自然科学版, 2011, 29(5):1-7.
[19]王晓轩,夏丽华,邓珊珊,等.基于RS和GIS的南海区基塘用地时空变化分析[J].资源与产业, 2011, 1313(4):55-60.
[20]麻锴,查东平,王宝健.基于高分一号宽幅多光谱影像的鄱阳湖湿地分类研究[J].测绘与空间地理信息, 2016, 3939(6):107-110.
[21]陈庆华,乔国军.基于WorldView 2遥感影像的湿地资源调查[J].测绘与空间地理信息, 2012,(S1):68-69.
[22]凌成星,鞠洪波,张怀清,等.基于Worldview-2八波段影像改进指数的湿地类型分类研究[J].林业科学研究, 2014, 2727(5):639-643.
[23]钟功甫.珠江三角洲的“桑基鱼塘”——一个水陆相互作用的人工生态系统[J].地理学报, 1980, 3535(3):200-209.
[24]钟功甫.珠江三角洲的“桑基焦塘”与“蔗基鱼塘”[J].地理学报, 1958, 2424(3):257-274.
[25]张军,涂丹,李国辉. 3S技术基础[M].北京:清华大学出版社,2013.
[26]汤国安,张友顺,刘咏梅,等.遥感数字图像处理[M].北京:科学出版社, 2004.
[27]姚琴风.多尺度城市道路高分辨率遥感影像提取方法研究[D].太原:太原理工大学, 2015.
[28]Ouyang Z T, Zhang M Q, Xie X, et al. A comparison of pixelbased and object-oriented approaches to VHR imagery for mapping saltmarsh plants[J]. Ecological Informatics, 2011, 6(2):136-146.
[29]Wolf A. Using WorldView 2 Vis-NIR MSI imagery to support land mapping and feature extraction using normalized difference index ratios[R]. Longmont, CO:DigitalGlobe, 2010.
[30]Breiman L. Random forests[J]. Machine Learning, 2001, 4545(1):5-32.
[31]Pearl J. Probabilisticreasoning in intelligent system:Networks of plausible inference[J]. Artificial Intelligence, 1991, 48(1):117-124.
[32]Heckerman D. Bayesian networks for data mining[J]. Data Mining and Knowledge Discovery, 1997, 11(1):79-119.
[33]闭小梅,闭瑞华. KNN算法综述[J].科技创新导报, 2009, 5(14):31.
[34]Zhu Y, Liu K, Liu L, et al. Exploring the potential of Worldview-2 red-edge band-based vegetation indices for estimation of mangrove leaf area index with machine learning algorithms[J]. Remote Sensing, 2017, 99(10):1060.
[35]Li X, Chen G, Liu J, et al. Effects of Rapid Eye imagery’s red-edge band and vegetation indices on land cover classification in an arid region[J]. Chinese Geographical Science, 2017, 27(5):827-835.
[36]Pu R, Landry S. A comparative analysis of high spatial resolution IKONOS and WorldView-2 imagery for mapping urban tree species[J]. Remote Sensing of Environment, 2012, 124(9):516-533.
[37]Haboudane D, Miller J R, Tremblay N, et al. Integrated narrowband vegetation indices for prediction of crop chlorophyll content for application to precision agriculture[J]. Remote Sensing of Environment, 2002, 8181(2):416-426.
[38]钟功甫.珠江三角洲桑基鱼塘生态系统若干问题研究[J].生态学杂志, 1982, 11(1):10-13.
[39]钟功甫.基塘系统的特征及其实践意义[J].地理科学, 1988, 8(1):12-17.
[2]钟功甫,蔡国雄.我国基(田)塘系统生态模式模式以珠江三角洲和长江三角洲为例[J].生态经济, 1987,(3):15-20.
[3]钟功甫.基塘系统的特征及其实践意义[J].地理科学, 1988, 8(1):12-17.
[4]钟功甫,王增骐,吴厚水.基塘系统的水陆相互作用[M].北京:科学出版社, 1993.
[5]聂呈荣,骆世明,章家恩,等.现代集约农业下基塘系统的退化与生态恢复[J].生态学报, 2003, 2323(9):1851-1860.
[6]丁疆华,温琰茂,舒强.基塘系统可持续发展现状、问题及对策研究[J].重庆环境科学, 2001, 2323(5):12-14.
[7]祝功武.珠三角人工地貌水文调节机制的今日思考[J].热带地理, 2012, 3232(4):378-384.
[8]袁兰,胡月明,范建彬.珠江三角洲农户整治基塘土地的意愿及影响因素研究[J].广东农业科学, 2015, 4242(7):180-186.
[9]刘青勇,王爱芹,张娜,等.基于调水调沙的黄河三角洲湿地生态修复技术[J].中国农村水利水电, 2016,(2):60-63.
[10]赵玉环.社会经济发展对珠江三角洲基塘系统的影响[J].仲恺农业技术学院学报, 2001, 1414(3):28-33.
[11]吴满强.历史时期佛山地区土地利用状况探析[D].广州:暨南大学, 2012.
[12]吴建新.明清民国顺德的基塘农业与经济转型[J].古今农业,2011,(1):96-104.
[13]杨丹.佛山市基塘景观破碎化动力机制及其模拟[D].南昌:东华理工大学, 2015.
[14]李梅,聂呈荣,龙兴.基塘系统生态环境质量评价指标体系的构建[J].农业环境科学学报, 2007, 2626(1):386-390.
[15]王晓锋,袁兴中,周李磊,等.三峡水库消落带景观基塘系统对地表径流氮磷拦截效益——以重庆开州区汉丰湖为例[J].三峡生态环境监测, 2017,(3):40-48.
[16]韩西丽,俞孔坚,李迪华,等.基塘-城市景观安全格局构建研究——以佛山市顺德区马岗片区为例[J].地域研究与开发,2008, 2727(5):107-110.
[17]林媚珍,冯荣光,纪少婷.中山市基塘农业模式演变及景观格局分析[J].广东农业科学, 2014, 4141(24):184-189.
[18]魏兴琥,王海,刁华涛.佛山市顺德区基塘农业模式的演变与发展潜力[J].佛山科学技术学院学报:自然科学版, 2011, 29(5):1-7.
[19]王晓轩,夏丽华,邓珊珊,等.基于RS和GIS的南海区基塘用地时空变化分析[J].资源与产业, 2011, 1313(4):55-60.
[20]麻锴,查东平,王宝健.基于高分一号宽幅多光谱影像的鄱阳湖湿地分类研究[J].测绘与空间地理信息, 2016, 3939(6):107-110.
[21]陈庆华,乔国军.基于WorldView 2遥感影像的湿地资源调查[J].测绘与空间地理信息, 2012,(S1):68-69.
[22]凌成星,鞠洪波,张怀清,等.基于Worldview-2八波段影像改进指数的湿地类型分类研究[J].林业科学研究, 2014, 2727(5):639-643.
[23]钟功甫.珠江三角洲的“桑基鱼塘”——一个水陆相互作用的人工生态系统[J].地理学报, 1980, 3535(3):200-209.
[24]钟功甫.珠江三角洲的“桑基焦塘”与“蔗基鱼塘”[J].地理学报, 1958, 2424(3):257-274.
[25]张军,涂丹,李国辉. 3S技术基础[M].北京:清华大学出版社,2013.
[26]汤国安,张友顺,刘咏梅,等.遥感数字图像处理[M].北京:科学出版社, 2004.
[27]姚琴风.多尺度城市道路高分辨率遥感影像提取方法研究[D].太原:太原理工大学, 2015.
[28]Ouyang Z T, Zhang M Q, Xie X, et al. A comparison of pixelbased and object-oriented approaches to VHR imagery for mapping saltmarsh plants[J]. Ecological Informatics, 2011, 6(2):136-146.
[29]Wolf A. Using WorldView 2 Vis-NIR MSI imagery to support land mapping and feature extraction using normalized difference index ratios[R]. Longmont, CO:DigitalGlobe, 2010.
[30]Breiman L. Random forests[J]. Machine Learning, 2001, 4545(1):5-32.
[31]Pearl J. Probabilisticreasoning in intelligent system:Networks of plausible inference[J]. Artificial Intelligence, 1991, 48(1):117-124.
[32]Heckerman D. Bayesian networks for data mining[J]. Data Mining and Knowledge Discovery, 1997, 11(1):79-119.
[33]闭小梅,闭瑞华. KNN算法综述[J].科技创新导报, 2009, 5(14):31.
[34]Zhu Y, Liu K, Liu L, et al. Exploring the potential of Worldview-2 red-edge band-based vegetation indices for estimation of mangrove leaf area index with machine learning algorithms[J]. Remote Sensing, 2017, 99(10):1060.
[35]Li X, Chen G, Liu J, et al. Effects of Rapid Eye imagery’s red-edge band and vegetation indices on land cover classification in an arid region[J]. Chinese Geographical Science, 2017, 27(5):827-835.
[36]Pu R, Landry S. A comparative analysis of high spatial resolution IKONOS and WorldView-2 imagery for mapping urban tree species[J]. Remote Sensing of Environment, 2012, 124(9):516-533.
[37]Haboudane D, Miller J R, Tremblay N, et al. Integrated narrowband vegetation indices for prediction of crop chlorophyll content for application to precision agriculture[J]. Remote Sensing of Environment, 2002, 8181(2):416-426.
[38]钟功甫.珠江三角洲桑基鱼塘生态系统若干问题研究[J].生态学杂志, 1982, 11(1):10-13.
[39]钟功甫.基塘系统的特征及其实践意义[J].地理科学, 1988, 8(1):12-17.