地球信息科学学报  2018 , 20 (2): 246-253 https://doi.org/10.12082/dqxxkx.2018.170360

遥感科学与应用技术

遥感土地覆被分类的空间尺度响应研究

徐凯健12, 田庆久1*, 杨闫君12, 徐念旭12

1. 南京大学国际地球系统科学研究所,南京 210023
2. 江苏省地理信息技术重点实验室,南京 210023

Response of Spatial Scale for Land Cover Classification of Remote Sensing

XU Kaijian12, TIAN Qingjiu1*, YANG Yanjun12, XU Nianxu12

1. International Institute for Earth System Science, Nanjing University, Nanjing 210023, China
2. Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology, Nanjing University, Nanjing 210023, China

通讯作者:  *通讯作者:田庆久(1964-),男,教授,博士,研究方向为高光谱遥感与遥感信息定量化。E-mail: tianqj@nju.edu.cn

收稿日期: 2017-08-3

修回日期:  2017-10-10

网络出版日期:  2018-03-02

版权声明:  2018 《地球信息科学学报》编辑部 《地球信息科学学报》编辑部 所有

基金资助:  国家重点研发计划重点专项(2017YFD0600903)国家科技重大专项(03-Y20A04-9001-17/18、30-Y20A29- 9003-15/17)

作者简介:

作者简介:徐凯健(1991-),男,博士生,研究方向为多尺度环境遥感应用。E-mail: phoenix-max@qq.com

展开

摘要

不同空间分辨率遥感影像对区域土地覆被类型识别精度的影响是目前土地资源遥感研究中的热点议题。本文基于准同步的卫星传感器影像,以福建省长汀县河田盆地为研究区,结合野外调查的实验样本,依次采用最大似然法(MLC)、支持向量机(SVM)和人工神经网络(ANN)3种分类器,分析土地覆被分类结果在中高空间尺度序列(1~50 m)下的变化响应特征。结果表明:不同空间尺度下的地物分类结果存在显著差异(P<0.05),其中总分类精度和Kappa系数均随影像分辨率的降低而先升高后降低,并于4 m分辨率处达到峰值,该结果与各类地物光谱反射率的空间尺度变化特征密切相关;而不同分类器对各空间尺度影像分类结果的影响程度差异较大(P<0.05),其中SVM的分类精度最优,MLC次之,ANN的结果较差。此外,伴随影像空间分辨率的降低,不同土地覆被类型面积提取结果的变化规律不同,导致同类地物在不同空间尺度下的提取结果出现较大差异,表明在使用多源分辨率遥感数据进行土地监测等相关研究时,其伴随的结果误差不容忽视。

关键词: 遥感分类 ; 分类器 ; 高分卫星 ; 空间分辨率 ; 尺度效应

Abstract

Classification based on remote sensing data has been widely applied in land cover mapping and the dynamic change monitoring research, of which the consequence is always strongly affected by spatial resolution of the used images. However, the response of multi-resolution images to remote sensing classification is still highly uncertain. Satellite observation could supply more and more multi-resolution images covering the same area at the same time and it would provide abundant data and technical support for study of remote sensing classification. In this study, the Hetian basin of Changting County in Fujian Province, was selected as a case to examine the performance of three typical classifiers (Maximum Likelihood Classification, MLC; Support Vector Machine, SVM; Artificial Neural Network, ANN). They were applied to satellite observations of temporal quasi-synchronous and multi-spatial resolution from medium to high spatial resolution (1~50 m) and we investigated the links between spatial resolution and remote sensing classification. Then, we also analyzed the spatial scale difference of spectrum reflectance, recognition accuracy and area extraction of five major land types (including arable land, forest land, water area, bare land and construction land) of the data with seven spatial resolution levels of 1, 2, 4, 8, 16, 30, and 50 m. They were supported with GF-1 PMS (pan and multi-spectra sensor), GF-2 PMS, GF-1 WFV (wide field view), Landsat-8 OLI (operational land imager) and GF-4 PMS data. 1845 recorded points observed in field survey were taken as training samples and validation samples. The results showed that along with the change of image spatial resolution from 1 to 50 m, (1) the mean spectra of bare land and construction land remained stable and no obvious changes occurred to water body, while the mean spectra of arable land and forest land decreased significantly when image resolution coarser than 4 m. The standard deviations of water body, bare land and construction land all increased constantly, while the standard deviations of arable land and forest land almost maintained stable. (2) The overall accuracy gradually decreased from 94.97±2.5% to 79.03±2.25% across the three classifiers, showing a gradually downward trend. Meanwhile, Kappa coefficient also gradually decreased from 0.93±0.03 to 0.72±0.03, which indicated that the accuracy of land cover classification was closely and sensitively related to the resolution of remote sensing images (P<0.05). (3) The calculation errors of the land types area would become larger as the image tend to be coarser, of which the area of arable land, bare land and construction land decreased significantly, the area of forest land increased, and the change of water body was not evident. The results above confirmed that when using multi-resolution images to generate land cover area or making area comparison refer to time serial data results, the errors from spatial database of various multi-scale could not be neglected, which would be more suitable to make the multi-scale transform for spatial effect correction. Our framework demonstrated the regular pattern of multiscale remote sensing classification and provided the prerequisite for scale conversion of classification products with different resolution in the future.

Keywords: classification ; classifier ; high-resolution satellite ; spatial resolution ; scale effect

0

PDF (1543KB) 元数据 多维度评价 相关文章 收藏文章

本文引用格式 导出 EndNote Ris Bibtex

徐凯健, 田庆久, 杨闫君, 徐念旭. 遥感土地覆被分类的空间尺度响应研究[J]. 地球信息科学学报, 2018, 20(2): 246-253 https://doi.org/10.12082/dqxxkx.2018.170360

XU Kaijian, TIAN Qingjiu, YANG Yanjun, XU Nianxu. Response of Spatial Scale for Land Cover Classification of Remote Sensing[J]. Journal of Geo-information Science, 2018, 20(2): 246-253 https://doi.org/10.12082/dqxxkx.2018.170360

1 引言

土地覆被类型的分类与识别几乎是所有地学研究中的基础工作,也是深入理解地表地理过程与人地关系演变的必要前提[1,2]。随着近几十年来遥感技术的快速发展,基于遥感数据的影像分类技术被广泛应用于土地类型制图和动态变化监测等研究 中[3,4],不同尺度卫星影像数据目前已成为该领域研究最主要的数据源。其中,中高空间分辨率(1~50 m)遥感图像以其较高的空间与时间分辨率、较低的获取成本,在资源环境监测中的应用潜力巨大[5,6]。特别是2013年以来,随着NASA新陆地卫星Landsat-8以及中国高分一号(GF-1)、高分二号(GF-2)与高分四号(GF-4)等对地观测卫星相继投入使用,使中高分辨率尺度多光谱遥感数据得到极大丰富,同时也为区域土地资源监测与应用提供了更多空间尺度方法的选择。

遥感土地覆被分类的精度和不确定性不仅受限于遥感光谱特征、数据处理方式等,还在很大程度上受到所使用遥感影像空间分辨率的影响,并使其研究结果产生较大不确定性[7,8]。其中,空间尺度问题是造成土地覆被识别复杂性的关键[9]。遥感中的尺度效应[10]是指由于遥感数据空间分辨率的变化而导致遥感数据在表达信息内容和数据分析产生差异的现象,并将导致地表特征提取[11,12]、空间格局[13]和生态参量反演[14]等结果发生改变。由于空间分辨率的差异而导致的土地覆被分类不确定性,将会进一步影响宏观尺度的地表环境变化检测,以及地表与生态、大气、水文等相关应用模型反演结果的准确性[15],因此其已成为一个亟待解决的科学问题。

目前,已有大量研究通过多种影像尺度上推手段,分析了不同空间分辨率数据地物分类结果的尺度效应[11-12,16-17],其中多数结果表明不同空间分辨率数据对不同地类识别和面积提取精度存在显著差异。然而,上述研究是利用重采样后得到的模拟影像而并非基于真实尺度的遥感影像进行相关实验的,而重采样过程不可避免地会导致影像不同程度的空间信息损失或光谱扭曲,使研究结论存在较大不确定性[18,19,20],导致目前空间尺度效应对土地覆被分类的影响仍缺乏明确结论。基于此,本文以长汀河田地区为研究区,开展遥感土地覆被分类的空间尺度响应研究。通过在中高分辨率范围内共选择7种连续空间尺度卫星影像(1、2、4、8、16、30和50 m),对研究区的土地类型依次进行3种常用分类器的遥感分类实验,以此分析不同空间分辨率的遥感数据对土地覆被类型识别及其面积提取精度的影响规律,评估不同遥感分类器的空间尺度识别效率差异,并探讨分类结果出现空间尺度差异的原因。研究对掌握区域土地覆被遥感分类结果的尺度效应有重要参考价值,亦可为多源遥感数据分类方法的选择和尺度转换提供依据。

2 数据源与研究方法

2.1 研究区概况

研究区位于福建省长汀县中部的河田盆地, 这里曾是该县最严重的水土流失区,也是目前中国 南方生态恢复的典型示范区。影像覆盖范围为(116°23~116°32′ E, 25°33~25°43′ N),土地面积约为304.72 km2图1)。该地区属亚热带季风气候,年平均气温18.2 ℃,年降雨量为1520~1700 mm,海拔范围在265~739 m之间,盆地内部分布有大面积花岗岩低缓丘陵,土壤类型以红壤为主。研究区内主要的土地覆被类型包括林地、耕地、水体、裸地与建设用地(各土地类型的覆盖面积比例分别为:耕地27.7%,林地60.5%,水体1%,裸地1.6%和建筑用地9.2%)。其中,林地大多分布在影像周边地势较高的区域;耕地和建设用地多数分布于盆地中央、以及山间的平缓地区;受人类活动影响,裸地多分布于荒山和部分建筑用地的外围区,水体主要包括盆地附近的湖泊及经该区西部和南部的3~4条主要河流。

图1   研究区地理位置

Fig. 1   Location of the study area

2.2 数据源

研究所用遥感数据分别来源于自中国资源卫星中心(http://www.cresda.com/CN/)和USGS网站(http://glovis.usgs.gov/)。为减少因季节性太阳高度角或植被物候导致的地物光谱误差,同时避免南方地区多云雾天气的影响,最终选择以下成像于夏秋季节的高质量数据,同时所有影像具备相似的波段信息(表1)。此外,课题组成员于2015年8月 中旬于研究区开展野外试验,获取研究区各类土地类型分布的随机样本共1845个,同时记录地面控制点30个。对获取的所有样本点进行筛选,标准为以地类的中心点为纯样本,并满足针对50 m分辨率影像的样本纯度,精选出代表性最优的455个作为训练样本,剩余部分作为验证样本。

表1   研究选用遥感影像信息

Tab. 1   The information of satellite images used in this study

传感器类型波段名称波长信息/μm空间分辨率/m成像日期
GF-2
PMS1
蓝/绿0.45~0.52/0.52~0.591/42015-08-27
2015-08-27
红/近红外0.63~0.69/0.77~0.891/4
GF-1
PMS2
蓝/绿0.45~0.52/0.52~0.592/82015-09-17
2015-09-17
红/近红外0.63~0.69/0.77~0.892/8
GF-1 WFV4蓝/绿0.45~0.52/0.52~0.59162015-08-03
2015-08-03
红/近红外0.63~0.69/0.77~0.8916
Landsat-8 OLI蓝/绿0.45~0.52/0.53~0.60302015-09-18
2015-09-18
红/近红外0.63~0.68/0.85~0.8930
GF-4
PMS
蓝/绿0.45~0.52/0.52~0.60502016-08-01
2016-08-01
红/近红外0.63~0.69/0.76~0.9050

新窗口打开

2.3 影像预处理

(1) 采用Gram-Schmidt变换,将GF-1与GF-2的全色波段融合以生成分辨率为1 m和2 m的多光谱影像。其中,Gram-Schmidt变换原理是将多光谱图像信息拟合为低空间分辨率全色影像作为第一分量,并参与生成其他正交分量,再用高分辨率的全色影像信息取代第一分量,进行Gram-Schmidt逆变换,即得到具有高空间分辨率和光谱分辨率的遥感影像。大量研究表明Gram-Schmidt变换不仅能增加空间信息,同时能较好地保持原光谱信息[21],使图像具备更高的细节对比度[22,23]。此处,采用信息熵与对比度2种指标对GF-1与GF-2影像融合前后的信息量和空间对比度进行评价。通过对融合前后的影像波段信息比较,发现各波段信息量得到明显增加,同时影像对比度也明显增强,使地物边缘轮廓更加清晰化(表2)。

表2   影像融合前后的波段信息变化

Tab. 2   Comparison of entropy and equivalent number of looks of images before and after fusion

传感器波段名称信息熵对比度
融合前融合后融合前融合后
GF-1 PMS蓝光1.592.0229.941.96
绿光1.622.0227.0743.43
红光1.652.0227.2340.51
近红外1.652.0124.5442.02
GF-2 PMS蓝光1.712.135.2549.08
绿光1.812.0932.0746.65
红光1.732.0935.0545.56
近红外1.892.0827.1544.16

新窗口打开

(2) 对所有影像依次完成正射校正、基于地面控制点的几何精校正处理,并以GF-2影像为基准对所有影像进行相对几何校正,使影像校正误差RMS均低于0.35个像元。同时将所有影像坐标投影系统进行统一(Transverse_Mercator),并设置相同的波段数量与顺序。

(3) 根据中国资源卫星应用中心和美国USGS官方网站提供的定标参数,对所有影像依次进行辐射定标、大气校正和辐射归一化处理,其中大气校正统一采用基于影像校正的QUAC算法[24]进行,以完成影像像元DN值至真实地表反射率的转换。

2.4 分类方法

研究选用最大似然分类(MLC)、支持向量机分类(SVM)和神经网络分类(ANN)3种遥感分类器开展研究区的空间尺度序列土地分类研究。其中,MLC计算速率快,同时分类精度较好[25];SVM是一种基于统计学习理论的机器学习算法,通过将输入特征映射到高维空间获取新特征,具有较高的识别精度[26];ANN通过模拟人脑神经系统的识别结构,具有高度非线性分类的识别能力[27]。上述3种分类器在不同空间分辨率影像分类中的应用效果仍不得而知。在分类器的参数设置中,对SVM选取最常用且最稳定的径向基函数(RBF)进行,并设置其核函数γ为1,惩罚系数C为100[28];通过综合相关文献[29]设置ANN的节点权重为0.9,权重调节速度为0.2,训练初试权重为0.9,训练步幅为0.9,隐藏层数为1,迭代次数为1000,传递函数采用logsitic进行。此外,除了影像的4个原始波段外,通过计算优选增加NDVI、NDWI和EVI指数作为分类器的特征变量进行计算。

NDVI=NIR-RNIR+R(1)

NDWI=G-NIRG+NIR(2)

EVI=2.5×(NIR-R)NIR+6.0×R-7.5×B-1(3)

为保证分类结果的可对比性,对不同分辨率的影像和不同分类方法均采用相同的训练样本和检验样本,对不同数据采用相同的分类器参数设置。同时,为避免混合像元,以影像地类斑块的中心点作为建立训练区,以确保最大空间尺度下样本像元的纯度。

2.5 统计分析

将验证样本随机等分为5组,并依次对分类结果采用混淆矩阵和Kappa统计进行精度评价[30]。采用ENVI 5.3软件进行影像光谱运算与分析, Origin 9.0软件进行数据图形的绘制,SPSS 20.0软件对不同空间尺度结果数据进行方差齐性检验,并对结果存在显著差异的数据组进行最小显著性差异法(LSD)检验。其中,方差齐性检验是统计学中判断不同样本的总体方差是否一致的常用算法,其先通过对样本总体特征做出假设,再通过抽样分析的统计推理,以判断该假设最终是否能被接受;LSD算法则一般用于检验实验组和对照组数据之间的差别是否达到显著性水平。

3 结果与讨论

3.1 不同土地类型波段反射率的空间尺度特征

不同地类光谱反射率的空间尺度变化是影响其分类结果的重要因素。图2显示了在4种主要空间分辨率尺度下,不同土地覆被类型像元光谱反射率均值与标准差的变化结果。从整体上看,不同地物的光谱曲线在形状上随空间分辨率降低的变化不大,但数值的变化相对明显。其中,耕地与林地类型的光谱曲线形态相近,反射率均值随分辨率下降呈现先短暂升高后持续降低的趋势,表明较高的分辨率有利于表征纯植被冠层的原始反射率,而随分辨率降低后像元内的阴影面积比重增大,导致其反射率降低。同时,耕地的各波段反射率均略高于林地,该结果与徐涵秋等[31]研究相似,这可能与林地的垂直结构差异导致其像元内阴影组分面积更多有关。此外,裸地和建筑用地的反射率变化特征相似,它们的光谱均值都在4 m分辨率以后达到稳定,同时低分辨率影像的反射率标准差明显高于较高分辨率影像,这与Meddens等[13]使用重采样影像得到的研究结果一致,表明伴随分辨率降低,混入该类像元的其它光谱组分数量增加,这将不利于其识别精度的提高。此外,水体的反射率均值随空间尺度变化基本保持稳定,但标准差明显增加。

图2   不同土地类型的多尺度光谱特征统计

Fig. 2   Statistics of multi-scale spectral characteristics of different land cover types

3.2 不同分类器与空间分辨率对土地覆被分类结果影响

图3可知,不同空间分辨率对影像分类结果的影响差异十分显著。其中,较高分辨率影像的总分类精度和Kappa系数均显著优于较低分辨率的影像结果(P<0.05),表明较高的空间分辨率对于区域土地类型的识别与分类具有明显优势。同时,在较高空间尺度范围内(如1~8 m),并非是影像分辨率越高其分类结果越好,其精度伴随分辨率降低呈现先升高后降低的趋势,其中SVM与MLC的分类精度和Kappa系数均在4 m分辨率处达到最高(P<0.05)。同时,3种分类器在2 m和8 m处的总分类结果较为接近(P>0.05),且高于1 m分辨率处的结果。这主要是由于研究区各土地覆被类型的最佳空间识别尺度出现在4 m附近,而受到“椒盐效应”影响,SVM与MLC在1 m和2 m分辨率处的分类精度均显著低于4 m分辨率处的结果(P<0.05)。此外,3种分类器在50 m分辨率处的总分类精度和Kappa系数均为最低(P<0.05),仅达到77.06%~81.47%和0.691-0.749,与其最佳分类结果相比分别下降了15.7%~18.3%和21.5%~24.2%。

图3   不同尺度下影像总分类精度与Kappa系数结果统计

Fig. 3   Overall accuracy and Kappa coefficients of classifications using different spatial resolutions

此外,本研究中16 m(WFV)与30 m(OLI)分辨率影像的分类结果相近,这主要是由于相比于其他研究影像(10 bit),OLI传感器具有更高的辐射分辨率(16 bit)、以及更窄的近红外反射率波段(表1),使其在理论上可以获得更高的地物细节识别能力,在一定程度上克服了其空间分辨率较低的影响。本研究结果表明,OLI与WFV多光谱影像(基于 4个相同波段)的分类精度在各级空间尺度上并无显著差异(P>0.05)。从3种常见分类方法在不同空间尺度下的分类结果来看,SVM在各级空间分辨率下得到的分类精度最高,其中在影像空间分辨率低于8 m时,SVM和MLC的总分类精度与Kappa系数均显著高于ANN(P<0.05);而当影像的空间分辨率高于16 m后,MLC与ANN之间的分类结果无差异显著(P>0.05),并且均显著低于SVM的分类结果(P<0.05)。研究中ANN分类结果整体上较差,这主要与该方法的最优参数类型选择、隐含层以及结点数目较难判定有关[27]

3.3 不同空间分辨率下的土地覆被类型面积提取 结果

目前遥感土地覆被类型面积提取结果的空间尺度效应仍不得而知。此处,采用分类精度最优的SVM分类器,考虑到空间分辨率最高的影像的提取效果一般最好[32],选择以1 m影像分类结果的土地覆被类型面积为基准,提取其余各分辨率影像分类结果的土地类型面积进行变化幅度对比,结果如 图4所示。统计结果表明:伴随影像空间分辨率的降低,耕地的提取面积会出现一定幅度的下降 (-17.2% ~ -2.7%);相反林地面积则保持着一个相对稳定的增加范围(8.5%~12%);裸地面积在1~4 m空间分辨率内出现短暂增加,但在分辨率低于8 m后出现大幅下降(-55.5%~17%);建设用地面积则呈现较大幅度的持续降低(-67.8%~-20.6%);水体面积的空间变化并无明显特征(-19.7%~17.3%)。

图4   不同空间尺度下的土地覆被类型面积变化

Fig. 4   Change of area information of landscape extracted from different spatial resolutions

研究表明,不同土地覆盖类型的面积提取结果与其斑块形状特征以及景观分布格局存在着密切关联,尤其是各地类之间的景观破碎化程度存在较大差异[20,33-34]。其中,在空间上呈狭长线状分布(如河流水系、道路交通线等)或散点状分布(如建设用地、池塘水库、山间田地等)的土地覆被类型斑块,其原始的景观破碎化程度较高,通常表现为图斑面积小、空间异质和不连续,因此在影像空间分辨率降低的过程中很容易融入到周围呈大面积集中分布的背景地物中(如连续植被、大面积湖泊)形成混合像元,并逐渐丢失原始地类的光谱属性。耕地、裸地、建筑用地以及部分水体皆属于此类,它们的面积比例均随着影像空间分辨率的下降出现了不同程度的缩减;相反,空间上呈大面积聚合且分布连续的林地面积则呈现稳定增长。此外,水体的面积变化无明显空间规律,这主要是因为在南方丘陵山区内大面积的湖泊与细长的河流水系通常会同时存在,伴随影像分辨率下降它们之间的空间面积变化往往存在此消彼长的现象(即破碎化程度较高的水体会丢失细小斑块,而大面积集中分布的水体面积则相应增加),因此其深层的变化规律还有待进一步研究揭示。综上所述,研究者在使用多尺度影像进行相关土地覆被类型提取、或基于多时间尺度典型地物面积变化的相关研究时,需要尤其注意由数据源空间分辨率差异而导致的不同土地类型面积提取的结果误差。

4 结论

本文采用中高空间分辨率多尺度卫星观测数据序列,结合不同常用遥感分类器,对研究区开展了典型土地覆被类型识别与面积提取的空间尺度响应研究。主要结论如下:

(1)随着遥感影像空间分辨率由高向低变化,研究区耕地和林地的各波段反射率均值呈现降低趋势,同时标准差变化不大;水体的光谱反射率均值基本稳定,但标准差明显增加;建设用地和裸地的反射率均值在影像分辨率低于4 m之后保持稳定,但标准差不断增加。表明耕地和林地在不同观测尺度下的光谱稳定性较高,相反水体、裸地和建设用地的光谱反射率差异伴随观测尺度的上升而增大。

(2)采用较高分辨率影像(<8 m)进行土地覆被类型的识别与分类效果最好,其中区域土地覆被的最佳空间识别尺度约为4 m。基于3种不同分类器获得的总分类精度和Kappa系数整体均随影像的空间分辨率降低而先升高后降低,并在50 m分辨率时到达最低,分别降低了15.7%~18.3%和21.5%~24.2%,表明不同空间分辨率的遥感数据对土地覆被类型识别精度影响较大。总体上看,SVM分类器的空间尺度识别效率优于MLC和ANN。

(3)伴随影像的空间尺度上升,其混合像元数量不断增加,同时分类精度下降,导致不同土地覆被类型面积提取结果的误差增大,并且产生明显的变化特征差异。其中,耕地、裸地和建筑用地的面积提取结果会呈现不同程度的低估;林地的面积提取结果则会相应出现高估;水体提取结果的空间变化规律不明显,仍有待于更加深入的研究。

未来有必要增加对更长尺度序列下的中-低空间分辨率土地覆被类型识别与变化特征的研究,以便更全面地掌握大尺度区域土地覆被识别的空间尺度变化规律。同时,后续研究将进一步探讨基于多尺度真实影像与经升尺度转换的模拟影像之间的光谱特征差别,及其对区域土地与环境监测结果的影响差异。

The authors have declared that no competing interests exist.


参考文献

[1] Gong P, Wang J, Yu L, et al.

Finer resolution observation and monitoring of global land cover: First mapping results with Landsat TM and ETM+ data

[J]. International Journal of Remote Sensing, 2013,34:2607-2654.

https://doi.org/10.1080/01431161.2012.748992      URL      [本文引用: 1]      摘要

We have produced the first 30 m resolution global land-cover maps using Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) data. We have classified over 6600 scenes of Landsat TM data after 2006, and over 2300 scenes of Landsat TM and ETM+ data before 2006, all selected from the green season. These images cover most of the world's land surface except Antarctica and Greenland. Most of these images came from the United States Geological Survey in level L1T (orthorectified). Four classifiers that were freely available were employed, including the conventional maximum likelihood classifier (MLC), J4.8 decision tree classifier, Random Forest (RF) classifier and support vector machine (SVM) classifier. A total of 91,433 training samples were collected by traversing each scene and finding the most representative and homogeneous samples. A total of 38,664 test samples were collected at preset, fixed locations based on a globally systematic unaligned sampling strategy. Two software tools, Global Analyst and Global Mapper developed by extending the functionality of Google Earth, were used in developing the training and test sample databases by referencing the Moderate Resolution Imaging Spectroradiometer enhanced vegetation index (MODIS EVI) time series for 2010 and high resolution images from Google Earth. A unique land-cover classification system was developed that can be crosswalked to the existing United Nations Food and Agriculture Organization (FAO) land-cover classification system as well as the International Geosphere-Biosphere Programme (IGBP) system. Using the four classification algorithms, we obtained the initial set of global land-cover maps. The SVM produced the highest overall classification accuracy (OCA) of 64.9% assessed with our test samples, with RF (59.8%), J4.8 (57.9%), and MLC (53.9%) ranked from the second to the fourth. We also estimated the OCAs using a subset of our test samples (8629) each of which represented a homogeneous area greater than 500 m鈥壝椻500 m. Using this subset, we found the OCA for the SVM to be 71.5%. As a consistent source for estimating the coverage of global land-cover types in the world, estimation from the test samples shows that only 6.90% of the world is planted for agricultural production. The total area of cropland is 11.51% if unplanted croplands are included. The forests, grasslands, and shrublands cover 28.35%, 13.37%, and 11.49% of the world, respectively. The impervious surface covers only 0.66% of the world. Inland waterbodies, barren lands, and snow and ice cover 3.56%, 16.51%, and 12.81% of the world, respectively.
[2] 赫英明,王汉杰,张洪峰.

遥感数据的土地覆盖分类

[J].解放军理工大学学报:自然科学版,2011,12(3):294-300.

URL      [本文引用: 1]     

[ He Y M, Wang H J, Zhang H F.

Land cover classification based on remote sensing data

[J]. Journal of PLA University of Science and Technology: Natural Science Edition, 2011,12(3):294-300. ]

URL      [本文引用: 1]     

[3] Sexton J O, Song X P, Feng M, et al.

Global, 30-m resolution continuous fields of tree cover: Landsat-based rescaling of MODIS vegetation continuous fields with lidar-based estimates of error

[J]. International Journal of Digital Earth, 2013,6(5):427-448.

https://doi.org/10.1080/17538947.2013.786146      URL      [本文引用: 1]      摘要

We developed a global, 30-m resolution dataset of percent tree cover by rescaling the 250-m MOderate-resolution Imaging Spectroradiometer (MODIS) Vegetation Continuous Fields (VCF) Tree Cover layer using circa- 2000 and 2005 Landsat images, incorporating the MODIS Cropland Layer to improve accuracy in agricultural areas. Resulting Landsat-based estimates maintained consistency with the MODIS VCF in both epochs (RMSE =8.6% in 2000 and 11.9% in 2005), but showed improved accuracy in agricultural areas and increased discrimination of small forest patches. Against lidar measurements, the Landsat-based estimates exhibited accuracy slightly less than that of the MODIS VCF (RMSE=16.8% for MODIS-based vs. 17.4% for Landsat-based estimates), but RMSE of Landsat estimates was 3.3 percentage points lower than that of the MODIS data in an agricultural region. The Landsat data retained the saturation artifact of the MODIS VCF at greater than or equal to 80% tree cover but showed greater potential for removal of errors through calibration to lidar, with post-calibration RMSE of 9.4% compared to 13.5% in MODIS estimates. Provided for free download at the Global Land Cover Facility (GLCF) website (www.landcover.org), the 30-m resolution GLCF tree cover dataset is the highest-resolution multi-temporal depiction of Earth''s tree cover available to the Earth science community.
[4] Zheng B J, Myint S W, Thenkabail P S, et al.

A support vector machine to identify irrigated crop types using time-series Landsat NDVI data

[J]. International Journal of Applied Earth Observation and Geoinformation, 2015,34:103-112.

https://doi.org/10.1016/j.jag.2014.07.002      URL      [本文引用: 1]      摘要

Site-specific information of crop types is required for many agro-environmental assessments. The study investigated the potential of support vector machines (SVMs) in discriminating various crop types in a complex cropping system in the Phoenix Active Management Area. We applied SVMs to Landsat time-series Normalized Difference Vegetation Index (NDVI) data using training datasets selected by two different approaches: stratified random approach and intelligent selection approach using local knowledge. The SVM models effectively classified nine major crop types with overall accuracies of >86% for both training datasets. Our results showed that the intelligent selection approach was able to reduce the training set size and achieved higher overall classification accuracy than the stratified random approach. The intelligent selection approach is particularly useful when the availability of reference data is limited and unbalanced among different classes. The study demonstrated the potential of utilizing multi-temporal Landsat imagery to systematically monitor crop types and cropping patterns over time in arid and semi-arid regions.
[5] Pu R L, Susan B.

Mapping seagrass coverage and spatial patterns with high spatial resolution IKONOS imagery

[J]. International Journal of Applied Earth Observation and Geoinformation, 2017,54:145-158.

https://doi.org/10.1016/j.jag.2016.09.011      URL      [本文引用: 1]      摘要

Seagrass habitats in subtidal coastal waters provide a variety of ecosystem functions and services and there is an increasing need to acquire information on spatial and temporal dynamics of this resource. Here, we explored the capability of IKONOS (IKO) data of high resolution (402m) for mapping seagrass cover [submerged aquatic vegetation (%SAV) cover] along the mid-western coast of Florida, USA. We also compared seagrass maps produced with IKO data with that obtained using the Landsat TM sensor with lower resolution (3002m). Both IKO and TM data, collected in October 2009, were preprocessed to calculate water depth invariant bands to normalize the effect of varying depth on bottom spectra recorded by the two satellite sensors and further the textural information was extracted from IKO data. Our results demonstrate that the high resolution IKO sensor produced a higher accuracy than the TM sensor in a three-class % SAV cover classification. Of note is that the OA of %SAV cover mapping at our study area created with IKO data was 5–20% higher than that from other studies published. We also examined the spatial distribution of seagrass over a spatial range of 4–24002m using the Ripley’s K function [L(d)] and IKO data that represented four different grain sizes [402m (one IKO pixel), 802m (202×022 IKO pixels), 1202m (302×023 IKO pixels), and 1602m (402×024 IKO pixels)] from moderate-dense seagrass cover along a set of six transects. The Ripley’s K metric repeatedly indicated that seagrass cover representing 402m02×02402m pixels displayed a dispersed (or slightly dispersed) pattern over distances of <4–802m, and a random or slightly clustered pattern of cover over 9–24002m. The spatial pattern of seagrass cover created with the three additional grain sizes (i.e., 202×022402m IKO pixels, 302×023402m IKO pixels, and 402×02402m IKO pixels) show a dispersed (or slightly dispersed) pattern across 4–3202m and a random or slightly clustered pattern across 33–24002m. Given the first report on using satellite observations to quantify seagrass spatial patterns at a spatial scale from 402m to 24002m, our novel analyses of moderate-dense SAV cover utilizing Ripley’s K function illustrate how data obtained from the IKO sensor revealed seagrass spatial information that would be undetected by the TM sensor with a 3002m pixel size. Use of the seagrass classification scheme here, along with data from the IKO sensor with enhanced resolution, offers an opportunity to synoptically record seagrass cover dynamics at both small and large spatial scales.
[6] 杨红卫,童小华.

中高分辨率遥感影像在农业中的应用现状

[J].农业工程学报,2012,28(24):138-149.

URL      Magsci      [本文引用: 1]      摘要

为了更加透彻地了解中高分辨率遥感影像技术在农业中的应用研究现状,本文通过对其应用研究现状进行了提炼和总结,并给以简洁的展望。明确了物种识别和分类、农作物面积估算与监测、作物养分和水分状况监测、农作物的叶面积指数、生物量和叶绿素含量监测、病虫害监测和预报、农业自然灾害的监测、农用地土地资源动态变化和监测、作物产量预测8个方面是中高分辨率遥感影像在农业工作中的应用重点,指出农业工作将更加精细、多种技术和方法的融合使用、统一的农业遥感影像数据库的完善、对当前遥感影像技术不足之处的弥补、标准农业植被光谱库的完善和健全的植被解译体系的完善6个方面将是未来一段时期内应用发展的重点和热点。本文可对遥感农业应用和研究工作提供一定参考。

[ Yang H W, Tong X H.

Application status of remote sensing images with middle and high resolution in agriculture

[J]. Transactions of the Chinese Society of Agricultural Engineering, 2012,28(24):138-149. ]

URL      Magsci      [本文引用: 1]      摘要

为了更加透彻地了解中高分辨率遥感影像技术在农业中的应用研究现状,本文通过对其应用研究现状进行了提炼和总结,并给以简洁的展望。明确了物种识别和分类、农作物面积估算与监测、作物养分和水分状况监测、农作物的叶面积指数、生物量和叶绿素含量监测、病虫害监测和预报、农业自然灾害的监测、农用地土地资源动态变化和监测、作物产量预测8个方面是中高分辨率遥感影像在农业工作中的应用重点,指出农业工作将更加精细、多种技术和方法的融合使用、统一的农业遥感影像数据库的完善、对当前遥感影像技术不足之处的弥补、标准农业植被光谱库的完善和健全的植被解译体系的完善6个方面将是未来一段时期内应用发展的重点和热点。本文可对遥感农业应用和研究工作提供一定参考。
[7] 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.

https://doi.org/10.1080/01431160600746456      URL      [本文引用: 1]     

[8] 李小文,王祎婷.

定量遥感尺度效应刍议

[J].地理学报,2013,68(9):5-11.

https://doi.org/10.11821/dlxb201309001      URL      [本文引用: 1]      摘要

目前定量遥感在面向国家需求、面向基础研究方面,存在海量遥感数据无法有效利用、定量遥感研究缺乏普适性和系统性、遥感应用难以再上新台阶等主要问题。因此,目前的遥感科学研究需要整合,需要向前迈出一大步。由于地表的异质性,我们如果只停留在演绎普适的机理模型在特定地点作个性化的处理,是不能适合地学的应用和研究的。必须在已有的反演和实验数据的基础上,用归纳的方法,总结出一些规律性的东西。走我国自然地理学"自上而下的演绎方法和自下而上归纳方法的结合"研究"尺度综合"的路子,在解决遥感科学核心问题"尺度效应"方面先搭建一个方法框架,同时建立几个开放的相关平台(如数据,反演,计算机模拟,等等),与有关学科的专家共享。

[ Li X W, Wang Y T.

Prospects on future developments of quantitative remote sensing

[J]. Acta Geographica Sinica, 2013,68(9):5-11. ]

https://doi.org/10.11821/dlxb201309001      URL      [本文引用: 1]      摘要

目前定量遥感在面向国家需求、面向基础研究方面,存在海量遥感数据无法有效利用、定量遥感研究缺乏普适性和系统性、遥感应用难以再上新台阶等主要问题。因此,目前的遥感科学研究需要整合,需要向前迈出一大步。由于地表的异质性,我们如果只停留在演绎普适的机理模型在特定地点作个性化的处理,是不能适合地学的应用和研究的。必须在已有的反演和实验数据的基础上,用归纳的方法,总结出一些规律性的东西。走我国自然地理学"自上而下的演绎方法和自下而上归纳方法的结合"研究"尺度综合"的路子,在解决遥感科学核心问题"尺度效应"方面先搭建一个方法框架,同时建立几个开放的相关平台(如数据,反演,计算机模拟,等等),与有关学科的专家共享。
[9] Woodcock C E, Strahler A H.

The factor of scale in remote sensing

[J]. Remote Sensing of Environment, 1987,21(3):311-332.

https://doi.org/10.1016/0034-4257(87)90015-0      URL      [本文引用: 1]      摘要

Thanks to such second- and third-generation sensor systems as Thematic Mapper, SPOT, and AVHRR, a user of digital satellite imagery for remote sensing of the earth's surface now has a choice of image scales ranging from 10 m to 1 km. The choice of an appropriate scale, or spatial resolution, for a particular application depends on several factors. These include the information desired about the ground scene, the analysis methods to be used to extract the information, and the spatial structure of the scene itself. A graph showing how the local variance of a digital image for a scene changes as the resolution-cell size changes can help in selecting an appropriate image scale. Such graphs are obtained by imaging the scene at fine resolution and then collapsing the image to successively coarser resolutions while calculating a measure of local variance. The local variance/resolution graphs for the forested, agricultural, and urban/suburban environments examined in this paper reveal the spatial structure of each type of scene, which is a function of the sizes and spatial relationships of the objects the scene contains. At the spatial resolutions of SPOT and Thematic Mapper imagery, local image variance is relatively high for forested and urban/suburban environments, suggesting that information-extracting techniques utilizing texture, context, and mixture modeling are appropriate for these sensor systems. In agricultural environments, local variance is low, and the more traditional classifiers are appropriate.
[10] Atkinson P M, Aplin P.

Spatial variation in land cover and choice of spatial resolution for remote sensing

[J]. International Journal of Remote Sensing, 2004,25(18):3687-3702.

https://doi.org/10.1080/01431160310001654383      URL      [本文引用: 1]      摘要

Prior to acquiring remotely sensed imagery with which to map land cover investigators may wish to select an appropriate spatial resolution. Previously, statistics such as the local variance and scale variance have been used to facilitate this goal. However, where such statistics vary locally over the region of interest, their use in selecting a single spatial resolution may be undermined. The variogram and scale variance (plotted as a function of spatial resolution) were predicted for airborne multispectral imagery with a spatial resolution of 4鈥塵 of St Albans, Hertfordshire, UK and of Arundel, Sussex, UK. The remotely sensed response in the red and near-infrared wavelengths was found to vary appreciably both within and between broad land categories (such as urban, agricultural and semi-natural areas). These differences mean that where the subject of interest is a general region rather than a specific feature or object the mean local variance or scale variance over that region may be unhelpful in selecting a single spatial resolution. Further, differences observed between the red and near-infrared wavelengths have implications for users who wish to select a single spatial resolution for multispectral imagery.
[11] Ballanti L, Blesius L, Hines E, et al.

Tree species classification using hyperspectral imagery: A comparison of two classifiers

[J]. Remote Sensing, 2016,8(6):445-462.

https://doi.org/10.3390/rs8060445      URL      [本文引用: 2]      摘要

The identification of tree species can provide a useful and efficient tool for forest managers for planning and monitoring purposes. Hyperspectral data provide sufficient spectral information to classify individual tree species. Two non-parametric classifiers, support vector machines (SVM) and random forest (RF), have resulted in high accuracies in previous classification studies. This research takes a comparative classification approach to examine the SVM and RF classifiers in the complex and heterogeneous forests of Muir Woods National Monument and Kent Creek Canyon in Marin County, California. The influence of object- or pixel-based training samples and segmentation size on the object-oriented classification is also explored. To reduce the data dimensionality, a minimum noise fraction transform was applied to the mosaicked hyperspectral image, resulting in the selection of 27 bands for the final classification. Each classifier was also assessed individually to identify any advantage related to an increase in training sample size or an increase in object segmentation size. All classifications resulted in overall accuracies above 90%. No difference was found between classifiers when using object-based training samples. SVM outperformed RF when additional training samples were used. An increase in training samples was also found to improve the individual performance of the SVM classifier.
[12] Roth K L, Roberts D A, Dennison P E, et al.

The impact of spatial resolution on the classification of plant species and functional types within imaging spectrometer data

[J]. Remote Sensing of Environment, 2015,171:45-57.

https://doi.org/10.1016/j.rse.2015.10.004      URL      [本文引用: 2]      摘要

Several upcoming hyperspectral satellite sensor missions (e.g., the Hyperspectral Infrared Imager and the Environmental Mapping and Analysis Program) will greatly expand the opportunities for researchers to use imaging spectroscopy data for discriminating and mapping plant species and plant functional types (PFTs; defined in this study as combinations of leaf-type, leaf/plant duration and life form). Accurate knowledge of the spatial distribution of dominant plant species and PFTs is highly valuable to many scientific and management goals, including improved parameterization of ecosystem process and climate models, better invasive species distribution monitoring and forecasting, quantification of human and natural disturbance and recovery processes, and evaluations of terrestrial vegetation response to climate change. Most often, species-level discrimination has been achieved using fine spatial resolution (≤022002m) airborne imagery, but currently proposed spaceborne imaging spectrometers will have coarser spatial resolution (~0230 to 6002m). In order to address the impact of coarser spatial resolutions on our ability to spectrally separate species and PFTs, we classified dominant species and PFTs in five contrasting ecosystems over a range of spatial resolutions. Study sites included a temperate broadleaf deciduous forest, a brackish tidal marsh, a mixed conifer/broadleaf montane forest, a temperate rainforest and a Mediterranean climate region encompassing grasslands, oak savanna, oak woodland and shrublands. Data were acquired by the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) over each site, and spatially aggregated to 20, 40 and 6002m resolutions. Canonical Discriminant Analysis (CDA) was used to classify species and PFTs at each site and across scales with overall accuracies ranging from 61 to 96% for species and 83–100% for PFTs. The results of this study show accuracy increases at coarser resolutions (≥022002m) across ecosystems, supporting the use of imaging spectroscopy data at spatial resolutions up to 6002m for the purpose of discriminating among plant species and PFTs. In four of the five study sites, the best accuracies were achieved at 4002m resolution. However, at coarser resolutions, some fine-scale species variation is lost and classes that occur only in small patches cannot be mapped. We also demonstrate that spectral libraries developed from fine spatial resolution imagery can be successfully applied as training data to accurately classify coarser resolution data over multiple ecosystems.
[13] Meddens A J H, Hicke J A, Vierling L A.

Evaluating the potential of multispectral imagery to map multiple stages of tree mortality

[J]. Remote Sensing of Environment, 2011,115:1632-1642.

https://doi.org/10.1016/j.rse.2011.02.018      URL      [本文引用: 2]      摘要

Insect outbreaks are major forest disturbances, causing tree mortality across millions of ha in North America. Resultant spatial and temporal patterns of tree mortality can profoundly affect ecosystem structure and function. In this study, we evaluated the classification accuracy of multispectral imagery at different spatial resolutions. We used four-band digital aerial imagery (30-cm spatial resolution and aggregated to coarser resolutions) acquired over lodgepole pine-dominated stands in central Colorado recently attacked by mountain pine beetle. Classes of interest included green trees and multiple stages of post-insect attack tree mortality, including dead trees with red needles (“red-attack”), dead trees without needles (“gray-attack”), and non-forest. The 30-cm resolution image facilitated delineation of trees located in the field, which were used in image classification. We employed a maximum likelihood classifier using the green band, Red–Green Index (RGI), and Normalized Difference Vegetation Index (NDVI). Pixel-level classification accuracies using this imagery were good (overall accuracy of 87%, kappa02=020.84), although misclassification occurred between a) sunlit crowns of live (green) trees and herbaceous vegetation, and b) sunlit crowns of gray- and red-attack trees and bare soil. We explored the capability of coarser resolution imagery, aggregated from the 30-cm resolution to 1.2, 2.4, and 4.202m, to improve classification accuracy. We found the highest accuracy at the 2.4-m resolution, where reduction in omission and commission errors and increases in overall accuracy (90%) and kappa (0.88) were achieved, and visual inspection indicated improved mapping. Pixels at this resolution included more shadow in forested regions than pixels in finer resolution imagery, thereby reducing forest canopy reflectance and allowing improved separation between forest and non-forest classes, yet were fine enough to resolve individual tree crowns better than the 4.2-m imagery. Our results illustrate that a classification of an image with a spatial resolution similar to the area of a tree crown outperforms that of finer and coarser resolution imagery for mapping tree mortality and non-forest classes. We also demonstrate that multispectral imagery can be used to separate multiple postoutbreak attack stages (i.e., red-attack and gray-attack) from other classes in the image.
[14] 王苗苗,周蕾,王绍强,.

空间分辨率对总初级生产力模拟结果差异的影响

[J].地理研究,2016,35(4):617-626.

URL      [本文引用: 1]     

[ Wang M M, Zhou L, Wang S Q, et al.

An analysis of the simulation differences of gross primary productivity resulting from the spatial resolution

[J]. Geographical Research, 2016,35(4):617-626. ]

URL      [本文引用: 1]     

[15] Yakar M, Yilmaz H M, Yurt K.

The effect of grid resolution in defining terrain surface

[J]. Experimental Techniques, 2010,34(6):23-29.

https://doi.org/10.1111/j.1747-1567.2009.00553.x      URL      [本文引用: 1]      摘要

ABSTRACT A study was conducted to investigate the effect of grid resolution in defining a terrain surface. The best comparison involved comparing a real value with an obtained value. Two artificial objects were used for defining the terrain in the investigations and volumes of these objects were measured in the laboratory. These volumes were taken as the real values and the most accurate volume was obtained from the best defined surface. These objects were evaluated using the photogrammetric method and points on surfaces of objects were measured. Volumes of objects were computed using different grid resolutions with Surfer 8.0 software. The same interpolation method was used in each application and parameters, such as point frequency, point distribution, and measurement accuracy remained constant. Photogrammetric techniques were used to conduct the investigations, as they allowed conversion of images of an object into a 3D model.
[16] 赵磊.

基于多源遥感数据的区域景观格局尺度效应

[J].遥感信息,2009,24(4):55-61.

URL      [本文引用: 1]     

[ Zhao L.

Scale effect of regional landscape patterns based on the multi-source remote sensing data

[J]. Remote Sensing Information, 2009,24(4):55-61. ]

URL      [本文引用: 1]     

[17] 韩鹏,龚健雅,李志林,.

遥感影像分类中的空间尺度选择方法研究

[J].遥感学报,2010,14(3):507-518.

URL      Magsci      [本文引用: 1]      摘要

提出了一种新的基于信息熵的空间尺度选择方法.该方法充分利用了遥感影像的多光谱信息.在这个方法中,信息熵被用于评价影像类别可分性的定量标准;另外影像的空间分布特征也被考虑.该方法与已有方法,即基于局部方差的方法、基于变异函数(Variogram)的方法、基于离散度的方法,进行了比较.TM和QuickBird两种影像被引入到评价中来.实验结果表明,本方法能够准确地确定两种实验影像的最优分类精度所对应的空间尺度.QuickBird影像采用了面向对象的分类方法进行实验,这表明本方法不仅适合于传统的分类方法,同时也适用于面向对象的方法.通过比较分析表明,本文方法明确优于已有各种方法.

[ Han P, Gong J Y, Li Z L, et al.

Selection of optimal scale in classification of remotely sensed images

[J]. Journal of Remote Sensing, 2010,14(3):507-518. ]

URL      Magsci      [本文引用: 1]      摘要

提出了一种新的基于信息熵的空间尺度选择方法.该方法充分利用了遥感影像的多光谱信息.在这个方法中,信息熵被用于评价影像类别可分性的定量标准;另外影像的空间分布特征也被考虑.该方法与已有方法,即基于局部方差的方法、基于变异函数(Variogram)的方法、基于离散度的方法,进行了比较.TM和QuickBird两种影像被引入到评价中来.实验结果表明,本方法能够准确地确定两种实验影像的最优分类精度所对应的空间尺度.QuickBird影像采用了面向对象的分类方法进行实验,这表明本方法不仅适合于传统的分类方法,同时也适用于面向对象的方法.通过比较分析表明,本文方法明确优于已有各种方法.
[18] Wang G X, Gertner G, Anderson A B.

Up-scaling methods based on variability-weighting and simulation for inferring spatial information across scales

[J]. International Journal of Remote Sensing, 2004,25(22):4961-4979.

https://doi.org/10.1080/01431160410001680428      URL      [本文引用: 1]      摘要

Appropriate up-scaling methods to infer spatial information from a finer to a coarser spatial resolution are required when remote sensing and geographical information systems (GIS) are used to generate multi-scale maps that are needed for agriculture, forestry, natural resources, environmental systems, and landscape ecology. The existing methods used in commercial GIS and image analysis packages such as Window Averaging (WA) often do not work well because of different limitations. In this study we developed and compared five widely used WA methods including three spatial variability-weighted methods and two simulation methods. These methods were assessed in a case study for aggregating and using Landsat Thematic Mapper (TM) images for mapping vegetation covers and for inferring a topographical factor related to soil erosion from finer to coarser resolutions. The results showed that the Beta Distribution Simulation (BDS) method was better than WA regardless of the distributions of the spatial data, while the Arithmetic Average Variability-Weighted method (AAVW) performed better than WA for normal distributions. BDS is flexible for variable distributions and AAVW is only suitable for normal distributions. Because of their simplicity, efficiency, and flexibility, it is expected that these two methods can be programmed into commercial GIS and image analysis packages.
[19] 韩鹏,龚健雅,李志林,.

遥感影像空间尺度上推方法的评价

[J].遥感学报,2008,12(6):964-971.

https://doi.org/10.3321/j.issn:1007-4619.2008.06.020      URL      Magsci      [本文引用: 1]      摘要

首先分析了几种常用的影像质量评价指标在遥感数据空间尺度上推方法评价中的不足,同时提出了已有的关于空间尺度上推方法在评价思路上的不妥之处.引入了空间分辨率和SSIM(structural SIMilarity)2个遥感影像质量评价指标,给出了新的评价思路,并在此基础上对5种遥感数据空间尺度转换方法进行了评价.在本实验中,空间分辨率和SSIM 2个遥感影像质量评价指标一致表明,Bilinear和Bicubic方法的结果影像能够更好地接近目标空间尺度下的影像特征.实验表明,依据新的评价思路,采用空间分辨率和SSIM 2个评价指标进行空间尺度上推方法表现出较强的有效性和优越性.

[ Han P, Gong J Y, Li Z L, et al.

Evaluation of the effect of aggregation methods for remote sensing images

[J]. Journal of Remote Sensing, 2008,12(6):964-971. ]

https://doi.org/10.3321/j.issn:1007-4619.2008.06.020      URL      Magsci      [本文引用: 1]      摘要

首先分析了几种常用的影像质量评价指标在遥感数据空间尺度上推方法评价中的不足,同时提出了已有的关于空间尺度上推方法在评价思路上的不妥之处.引入了空间分辨率和SSIM(structural SIMilarity)2个遥感影像质量评价指标,给出了新的评价思路,并在此基础上对5种遥感数据空间尺度转换方法进行了评价.在本实验中,空间分辨率和SSIM 2个遥感影像质量评价指标一致表明,Bilinear和Bicubic方法的结果影像能够更好地接近目标空间尺度下的影像特征.实验表明,依据新的评价思路,采用空间分辨率和SSIM 2个评价指标进行空间尺度上推方法表现出较强的有效性和优越性.
[20] 胡云锋,徐芝英,刘越,.

空间尺度上推方法的精度评价——以内蒙古锡林郭勒盟土地利用数据为例

[J].地理研究,2012,31(11):1961-1972.

https://doi.org/10.11821/yj2012110004      URL      [本文引用: 2]      摘要

不同的空间尺度上推方法会导致不同程度的信息丢失、信息歪曲等后果;但目前少有研究在较长的尺度序列上、对不同尺度上推方法所得成果开展精度分析。本研究首先提出尺度上推方法精度评价的三个准则,即:保持土地类型构成特征、保持土地面积精度、保持土地空间分布格局和斑块形态;继而使用格点中心值、最大面积斑块、最大聚合面积斑块等3种尺度上推方法,配合100m~50km土地利用数据开展尺度转化实验;最后基于上述评价准则和尺度上推实验所得的系列输出结果,分析了不同尺度上推方法的精度。研究表明:(1)格点中心值上推方法能更好地保留区域土地类型构成、土地面积精度等特征;(2)锡林郭勒盟地区土地研究的适宜尺度应小于10km,最大不应超过30km;(3)在尺度上推过程中,土地斑块的平均面积、形态以及空间分布格局对尺度上推成果精度有着重要影响。

[ Hu Y F, Xu Z Y, Liu Y, et al.

Accuracy analysis of up-scaling data: A case study with land use data in Xilin Guole of Inner Mongolia, China

[J]. Geographical Research, 2012,31(11):1961-1972. ]

https://doi.org/10.11821/yj2012110004      URL      [本文引用: 2]      摘要

不同的空间尺度上推方法会导致不同程度的信息丢失、信息歪曲等后果;但目前少有研究在较长的尺度序列上、对不同尺度上推方法所得成果开展精度分析。本研究首先提出尺度上推方法精度评价的三个准则,即:保持土地类型构成特征、保持土地面积精度、保持土地空间分布格局和斑块形态;继而使用格点中心值、最大面积斑块、最大聚合面积斑块等3种尺度上推方法,配合100m~50km土地利用数据开展尺度转化实验;最后基于上述评价准则和尺度上推实验所得的系列输出结果,分析了不同尺度上推方法的精度。研究表明:(1)格点中心值上推方法能更好地保留区域土地类型构成、土地面积精度等特征;(2)锡林郭勒盟地区土地研究的适宜尺度应小于10km,最大不应超过30km;(3)在尺度上推过程中,土地斑块的平均面积、形态以及空间分布格局对尺度上推成果精度有着重要影响。
[21] 孙攀,董玉森,陈伟涛,.

高分二号卫星影像融合及质量评价

[J].国土资源遥感,2016,28(4):108-113.

https://doi.org/10.6046/gtzyyg.2016.04.17      URL      Magsci      [本文引用: 1]      摘要

<p>高分二号卫星(GF-2)是我国自主研制的首颗空间分辨率优于1 m的民用光学遥感卫星,配备有0.81 m空间分辨率的全色相机和3.24 m空间分辨率的多光谱相机。对比分析适合GF-2影像的融合方法对于提高其应用效果与扩大应用领域具有实际意义。针对东北地区2014年11月22日和27日成像的GF-2影像,分别采用主成分分析(principal component analysis,PCA)、GS (Gram-Schmidt)变换、modified-HIS (intensity hue saturation)变换、高通滤波方法(high pass filter,HPF)和超球体色彩空间变换(hyperspherical color space resolution merge,HCS)等5种融合方法对多光谱和全色数据进行融合。并对5种融合影像进行质量评价,首先采用目视分析方法进行定性评价,其次采用信息熵、平均梯度、相关系数和光谱扭曲度等统计学指标进行客观定量评价,最后对融合影像进行地物分类。结果表明,HCS与GS变换融合影像无论是在视觉还是在地物分类应用上都具有较好的效果,且没有波段数的限制,最适合GF-2影像融合;HPF方法对空间细节信息的增强仅次于HCS变换,但是其光谱保真度效果最差;PCA和modified-IHS变换融合效果比较适中,可以作为GF-2影像融合的候补方法。</p>

[ Sun P, Dong Y S, Chen W T, et al.

Research on fusion of GF-2 imagery and quality evaluation

[J]. Remote Sensing for Land and Resources, 2016,28(4):108-113. ]

https://doi.org/10.6046/gtzyyg.2016.04.17      URL      Magsci      [本文引用: 1]      摘要

<p>高分二号卫星(GF-2)是我国自主研制的首颗空间分辨率优于1 m的民用光学遥感卫星,配备有0.81 m空间分辨率的全色相机和3.24 m空间分辨率的多光谱相机。对比分析适合GF-2影像的融合方法对于提高其应用效果与扩大应用领域具有实际意义。针对东北地区2014年11月22日和27日成像的GF-2影像,分别采用主成分分析(principal component analysis,PCA)、GS (Gram-Schmidt)变换、modified-HIS (intensity hue saturation)变换、高通滤波方法(high pass filter,HPF)和超球体色彩空间变换(hyperspherical color space resolution merge,HCS)等5种融合方法对多光谱和全色数据进行融合。并对5种融合影像进行质量评价,首先采用目视分析方法进行定性评价,其次采用信息熵、平均梯度、相关系数和光谱扭曲度等统计学指标进行客观定量评价,最后对融合影像进行地物分类。结果表明,HCS与GS变换融合影像无论是在视觉还是在地物分类应用上都具有较好的效果,且没有波段数的限制,最适合GF-2影像融合;HPF方法对空间细节信息的增强仅次于HCS变换,但是其光谱保真度效果最差;PCA和modified-IHS变换融合效果比较适中,可以作为GF-2影像融合的候补方法。</p>
[22] Nikolakopoulos K G.

Spatial Resolution enhancement of hyperion hyperspectral data

[J]. IEEE Hyperspectral Image and Signal Processing: Evolution in Remote Sensing, 2009,8:1-4.

https://doi.org/10.1109/WHISPERS.2009.5288996      URL      [本文引用: 1]      摘要

In this study eight fusion techniques and more especially the Ehlers, Gram-Schmidt, high pass filter, local mean matching (LMM), local mean and variance matching (LMVM), modified IHS (Modihs), Pansharp and PCA, were used for the fusion of Hyperion hyperspectral data with ALI panchromatic data. Both sensors are part of the Earth-Observing 1 satellite. The panchromatic data have a spatial resolution of 10 m while the hyperspectral data have a spatial resolution of 30 m. All the fusion techniques are designed for use with classical multispectral data. Thus, it is quite interesting to investigate the assessment of the common used fusion algorithms with the hyperspectral data.
[23] Guo J H, Yang F, Tan H, et al.

Shadow extraction from high-resolution remote sensing images based on gram-schmidt orthogonalization in lab space

[C]. 3rd International Symposium of Space Optical Instruments and Applications, Springer Proceedings in Physics, 2017,192:321-328.

[本文引用: 1]     

[24] Bernstein L S, Jin X M, Gregor B, et al.

Quick atmospheric correction code: Algorithm description and recent upgrades

[J]. Optical Engineering, 2012,51(11):1719.

https://doi.org/10.1117/1.OE.51.11.111719      URL      [本文引用: 1]      摘要

The quick atmospheric correction (QUAC) code performs atmospheric correction on multi- and hyperspectral imagery spanning all or part of the visible and near infrared-short wave infrared spectral range, 藴400-2500 nm. It utilizes an in-scene approach, requiring only approximate specification of sensor band locations (i.e., central wavelengths) and their radiometric calibration; no additional metadata is required. Because QUAC does not involve first principles radiative-transfer calculations, it is significantly faster than physics-based methods; however, it is also more approximate. We present a detailed description of the QUAC algorithm, highlighting recent accuracy improvements. Example results for several multi-and hyperspectral data sets are presented, and comparisons are made to more rigorous correction approaches.
[25] Hicke J A, Logan J A.

Mapping whitebark pine mortality caused by a mountain pine beetle outbreak with high spatial resolution satellite imagery

[J]. International Journal of Remote Sensing, 2009,30:4427-4441.

https://doi.org/10.1080/01431160802566439      URL      [本文引用: 1]      摘要

Insect outbreaks cause significant tree mortality across western North America, including in high-elevation whitebark pine forests. These forests are under several threats, which include attack by insects and white pine blister rust, as well as conversion to other tree species as a result of fire suppression. Mapping tree mortality is critical to determining the status of whitebark pine as a species. Satellite remote sensing builds upon existing aerial surveys by using objective, repeatable methods that can result in high spatial resolution monitoring. Past studies concentrated on level terrain and only forest vegetation type. The objective of this study was to develop a means of classifying whitebark pine mortality caused by a mountain pine beetle infestation in rugged, remote terrain using high spatial resolution satellite imagery. We overcame three challenges of mapping mortality in this mountainous region: (1) separating non-vegetated cover types, green and brown herbaceous cover, green (live) tree cover, and red-attack (dead) tree cover; (2) variations in illumination as a result of variations in slope and aspect related to the mountainous terrain of the study site; and (3) the difficulty of georegistering the imagery for use in comparing field measurements. Quickbird multi-spectral imagery (2.4聽m spatial resolution) was used, together with a maximum likelihood classification method, to classify vegetation cover types over a 6400 ha area. To train the classifier, we selected pixels in each cover class from the imagery guided by our knowledge of the study site. Variables used in the maximum likelihood classifier included the ratio of red reflectance to green reflectance as well as green reflectance. These variables were stratified by solar incidence angle to account for illumination variability. We evaluated the results of the classified image using a reserved set of image-derived class members and field measurements of live and dead trees. Classification results yielded high overall accuracy (86% and 91% using image-derived class members and field measurements respectively) and kappa statistics (0.82 and 0.82) and low commission (0.9% and 1.5%) and omission (6.5% and 15.9%) errors for the red-attack tree class. Across the scene, 700 ha or 31% of the forest was identified as in the red-attack stage. Severity (percent mortality by canopy cover) varied from nearly 100% for some areas to regions with little mortality. These results suggest that high spatial resolution satellite imagery can provide valuable information for mapping and monitoring tree mortality even in rugged, mountainous terrain.
[26] Bruzzone L, Persello C.

A novel context-sensitive semisupervised SVM classifier robust to mislabeled training samples

[J]. IEEE Transactions on Geoscience & Remote Sensing, 2009,47:2142-2154.

https://doi.org/10.1109/TGRS.2008.2011983      URL      [本文引用: 1]      摘要

This paper presents a novel context-sensitive semisupervised support vector machine (CS4VM) classifier, which is aimed at addressing classification problems where the available training set is not fully reliable, i.e., some labeled samples may be associated to the wrong information class (mislabeled patterns). Unlike standard context-sensitive methods, the proposed CS4VM classifier exploits the contextual information of the pixels belonging to the neighborhood system of each training sample in the learning phase to improve the robustness to possible mislabeled training patterns. This is achieved according to both the design of a semisupervised procedure and the definition of a novel contextual term in the cost function associated with the learning of the classifier. In order to assess the effectiveness of the proposed CS4VM and to understand the impact of the addressed problem in real applications, we also present an extensive experimental analysis carried out on training sets that include different percentages of mislabeled patterns having different distributions on the classes. In the analysis, we also study the robustness to mislabeled training patterns of some widely used supervised and semisupervised classification algorithms (i.e., conventional support vector machine (SVM), progressive semisupervised SVM, maximum likelihood, and k-nearest neighbor). Results obtained on a very high resolution image and on a medium resolution image confirm both the robustness and the effectiveness of the proposed CS4VM with respect to standard classification algorithms and allow us to derive interesting conclusions on the effects of mislabeled patterns on different classifiers.
[27] Shao Y, Lunetta R S.

Comparison of support vector machine, neural network, and CART algorithms for the land-cover classification using limited training data points

[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2012,70:78-87.

https://doi.org/10.1016/j.isprsjprs.2012.04.001      URL      [本文引用: 2]      摘要

Support vector machine (SVM) was applied for land-cover characterization using MODIS time-series data. Classification performance was examined with respect to training sample size, sample variability, and landscape homogeneity (purity). The results were compared to two conventional nonparametric image classification algorithms: multilayer perceptron neural networks (NN) and classification and regression trees (CART). For 2001 MODIS time-series data, SVM generated overall accuracies ranging from 77% to 80% for training sample sizes from 20 to 800 pixels per class, compared to 67–76% and 62–73% for NN and CART, respectively. These results indicated that SVM’s had superior generalization capability, particularly with respect to small training sample sizes. There was also less variability of SVM performance when classification trials were repeated using different training sets. Additionally, classification accuracies were directly related to sample homogeneity/heterogeneity. The overall accuracies for the SVM algorithm were 91% (Kappa02=020.77) and 64% (Kappa02=020.34) for homogeneous and heterogeneous pixels, respectively. The inclusion of heterogeneous pixels in the training sample did not increase overall accuracies. Also, the SVM performance was examined for the classification of multiple year MODIS time-series data at annual intervals. Finally, using only the SVM output values, a method was developed to directly classify pixel purity. Approximately 65% of pixels within the Albemarle–Pamlico Basin study area were labeled as “functionally homogeneous” with an overall classification accuracy of 91% (Kappa02=020.79). The results indicated a high potential for regional scale operational land-cover characterization applications.
[28] Murthy C S, Raju P V,Badrinath K V S.

Classification of wheat crop with multi-temporal images: Performance of maximum likelihood and artificial neural networks

[J]. International Journal of Remote Sensing, 2003,23:871-890.

https://doi.org/10.1080/0143116031000070490      URL      [本文引用: 1]      摘要

The need for multi-temporal data analysis for delineation of wheat crop has been demonstrated first. It is found that Maximum Likelihood Classification (MLC) with the composite data of multi-temporal images is limited by the problem of large null set containing crop pixels. Therefore, for effective classification of multi-temporal images, two approaches are evaluated: (1) MLC with different strategies鈥攕equential MLC (s_MLC), MLC with Principal Components (pca_MLC) and iterative MLC (i_MLC); and (2) Artificial Neural Networks (ANN) with back-propagation method. These classifiers were applied on multi-temporal Indian Remote Sensing satellite (IRS)-1B images to classify wheat crop in two areas of India, one with dominant wheat and the other with less dominant wheat cultivation. Among the three strategies of MLC, i_MLC has resulted in relatively better classification of wheat. However, the result of ANN classification is superior to that of i_MLC with respect to the correctness of labelling of wheat pixels. The performance of ANN is proved to be better, in both the situations of dominant wheat and less dominant wheat cultivation.
[29] 李颖,李耀辉,王金鑫,.

SVM和ANN在多光谱遥感影像分类中的比较研究

[J].海洋测绘,2016,36(5):19-22.

https://doi.org/10.3969/j.issn.1671-3044.2016.05.005      URL      [本文引用: 1]      摘要

首先利用支持向量机(SVM)和人工神经网络(ANN)对Landsat 80LI多光谱影像进行基于光谱信息的土地利用监督分类;然后,对多波段进行主成分变换,提取第一主成分的主要纹理信息,与光谱信息一起进行融合光谱和纹理信息的SVM和ANN影像监督分类.对比分析发现:对中原地区,SVM是Landsat 8多光谱遥感影像分类的较优方法,尤其适用于农业用地信息提取;光谱分类即可达到较高精度,纹理信息对提高分类精度的作用十分有限.

[ Li Y, Li Y H, Wang J X, et al.

A comparative study of SVM and ANN in multispectral image classification

[J]. Hydrographic Surveying and Charting, 2016,36(5):19-22. ]

https://doi.org/10.3969/j.issn.1671-3044.2016.05.005      URL      [本文引用: 1]      摘要

首先利用支持向量机(SVM)和人工神经网络(ANN)对Landsat 80LI多光谱影像进行基于光谱信息的土地利用监督分类;然后,对多波段进行主成分变换,提取第一主成分的主要纹理信息,与光谱信息一起进行融合光谱和纹理信息的SVM和ANN影像监督分类.对比分析发现:对中原地区,SVM是Landsat 8多光谱遥感影像分类的较优方法,尤其适用于农业用地信息提取;光谱分类即可达到较高精度,纹理信息对提高分类精度的作用十分有限.
[30] Lucas I F, Frans J M.

Accuracy assessment of satellite derived land-cover data: A review

[J]. Photogrammetric Engineering & Remote Sensing, 1994,60(4):410-432.

https://doi.org/10.1016/0924-2716(94)90066-3      URL      [本文引用: 1]      摘要

Accuracy assessment of land-cover classifications derived from remote sensing data has been recognized as a valuable tool in judging the fitness of these data for a particular application. Recent research initiatives in the area of spatial data accuracy and integration of remote sensing data in geographic information systems have revived the discussion on accuracy assessment. This article aims at contributing to this discussion by means of a review based on a division into positional and thematic accuracy. An important observation is that there are a limited number of methods for assessing data accuracy. However, the applied definitions differ very much from author to author, especially in the assessment of thermatic accuracy. Accuracy assessment mostly yields one single measure such as root-mean-square error or proportion of pixels correctly classified. These single measures do not give sufficient information and they can be based on statistically or methodologically non-valid methods. Therefore, not a single measure but also the total process of assessing these measures should explicitly be reported.
[31] 徐涵秋,刘智才,郭燕滨.

GF-1 PMS1与ZY-3 MUX传感器NDVI数据的对比分析

[J].农业工程学报,2016,32(8):148-154.

URL      Magsci      [本文引用: 1]      摘要

中国民用对地观测卫星在近10 a得到迅速发展,2012年和2013年相继发射的ZY-3和GF-1遥感卫星已成为中国现阶段主要应用的高分影像卫星,但二者对地观测能力是否相同并不清楚。因此,基于2对同日过空的GF-1 PMS1和ZY-3 MUX影像对,利用归一化植被指数NDVI对二者的植被观测能力进行对比。结果表明,GF-1 PMS1和ZY-3 MUX的植被观测能力虽然很接近,但也存在一定的差异。主要表现在ZY-3 MUX植被指数NDVI的信息量和信号总体强于GF-1 PMS1,后者的低估幅度为可达?3%;但随着NDVI的增强,GF-1 PMS1的低估会逐渐减少,在NDVI的高值区甚至可超过ZY-3 MUX。由于二者之间存在差异,因此它们如要应用于同一项目,建议要进行数据转换,以确保结果的准确对比。分析表明,这2种传感器数据之间的差异是二者在光谱响应函数、空间分辨率以及定标精度等方面的差异引起的。

[ Xu H Q, Liu Z C, Guo Y B.

Comparison of NDVI data between GF-1 PMS1 and ZY-3 MUX sensors

[J]. Transactions of the Chinese Society of Agricultural Engineering, 2016,32(8):148-154. ]

URL      Magsci      [本文引用: 1]      摘要

中国民用对地观测卫星在近10 a得到迅速发展,2012年和2013年相继发射的ZY-3和GF-1遥感卫星已成为中国现阶段主要应用的高分影像卫星,但二者对地观测能力是否相同并不清楚。因此,基于2对同日过空的GF-1 PMS1和ZY-3 MUX影像对,利用归一化植被指数NDVI对二者的植被观测能力进行对比。结果表明,GF-1 PMS1和ZY-3 MUX的植被观测能力虽然很接近,但也存在一定的差异。主要表现在ZY-3 MUX植被指数NDVI的信息量和信号总体强于GF-1 PMS1,后者的低估幅度为可达?3%;但随着NDVI的增强,GF-1 PMS1的低估会逐渐减少,在NDVI的高值区甚至可超过ZY-3 MUX。由于二者之间存在差异,因此它们如要应用于同一项目,建议要进行数据转换,以确保结果的准确对比。分析表明,这2种传感器数据之间的差异是二者在光谱响应函数、空间分辨率以及定标精度等方面的差异引起的。
[32] 王利民,刘佳,高建孟,.

冬小麦面积遥感识别精度与空间分辨率的关系

[J].农业工程学报,2016,32(23):152-160.

https://doi.org/10.11975/j.issn.1002-6819.2016.23.021      URL      [本文引用: 1]      摘要

不同空间分辨率农作物面积识别精度是农情遥感监测数据源选择的依据。该文采用WFV(wide field view)、MODIS(moderate-resolution imaging spectroradiometer)、OLI(operational land imager)、Google Earth影像,在天津市武清区选择了12 km×14 km的冬小麦种植区作为研究区域,采用目视识别的方法,分析了2、5、10、15、30、100、250 m共7个空间分辨率尺度下冬小麦面积识别精度与遥感数据分辨率、农田景观破碎度之间的关系。结果表明,随着空间分辨率由2 m变化到250 m,冬小麦面积识别的总体精度逐步由98.6%降低到70.1%,精度降低28.5%;面积数量比例由5.5%扩大到110.6%,误差增加105.1个百分点;面积精度呈明显下降趋势,数量误差呈明显增加趋势,数量误差的增加速度高于精度下降的趋势。高、中、低3个景观破碎度条件下,随着分辨率由2 m降低到250 m,作物识别精度分别降低了72.8、63.2和47.0个百分点,破碎度的增加导致面积识别精度下降速度更快;同等分辨率下,破碎度越高的地区面积识别精度越低。像元内冬小麦占比与可识别能力密切相关,像元占比达到45.0%以上时才能够被正确识别为冬小麦类型,像元尺度降低导致细小斑块丢失是造成面积识别与数量精度降低的主要原因。像元空间分辨率越高,冬小麦像元的光谱一致性越强,越有利于冬小麦分类精度的提高。针对农情遥感监测业务运行的需要,上述研究结果可以作为区域范围不同用户精度要求前提下遥感数据源选择的依据。

[ Wang L M, Liu J, Gao J M, et al.

Relationship between accuracy of remote sensing identification of winter wheat area and spatial resolution

[J]. Transactions of the Chinese Society of Agricultural Engineering, 2016,32(23):152-160. ]

https://doi.org/10.11975/j.issn.1002-6819.2016.23.021      URL      [本文引用: 1]      摘要

不同空间分辨率农作物面积识别精度是农情遥感监测数据源选择的依据。该文采用WFV(wide field view)、MODIS(moderate-resolution imaging spectroradiometer)、OLI(operational land imager)、Google Earth影像,在天津市武清区选择了12 km×14 km的冬小麦种植区作为研究区域,采用目视识别的方法,分析了2、5、10、15、30、100、250 m共7个空间分辨率尺度下冬小麦面积识别精度与遥感数据分辨率、农田景观破碎度之间的关系。结果表明,随着空间分辨率由2 m变化到250 m,冬小麦面积识别的总体精度逐步由98.6%降低到70.1%,精度降低28.5%;面积数量比例由5.5%扩大到110.6%,误差增加105.1个百分点;面积精度呈明显下降趋势,数量误差呈明显增加趋势,数量误差的增加速度高于精度下降的趋势。高、中、低3个景观破碎度条件下,随着分辨率由2 m降低到250 m,作物识别精度分别降低了72.8、63.2和47.0个百分点,破碎度的增加导致面积识别精度下降速度更快;同等分辨率下,破碎度越高的地区面积识别精度越低。像元内冬小麦占比与可识别能力密切相关,像元占比达到45.0%以上时才能够被正确识别为冬小麦类型,像元尺度降低导致细小斑块丢失是造成面积识别与数量精度降低的主要原因。像元空间分辨率越高,冬小麦像元的光谱一致性越强,越有利于冬小麦分类精度的提高。针对农情遥感监测业务运行的需要,上述研究结果可以作为区域范围不同用户精度要求前提下遥感数据源选择的依据。
[33] Li N, Xie G D, Zhou D, et al.

Remote sensing classification of Marsh Wetland with different resolution images

[J]. Journal of Resources and Ecology, 2016,7(2):107-114.

https://doi.org/10.5814/j.issn.1674-764x.2016.02.005      URL      [本文引用: 1]      摘要

Successful biological monitoring depends on judicious classification. An attempt has been made to provide an overview of important characteristics of marsh wetland. Classification was used to describe ecosystems and land cover patterns. Different spatial resolution images show different landscape characteristics. Several classification images were used to map and monitor wetland ecosystems of Honghe National Nature Reserve(HNNR) at a plant community scale. HNNR is a typical inland wetland and fresh water ecosystem in the North Temperate Zone. SPOT-5 10 m 脳 10 m, 20 m 脳 20 m, and 30 m 脳 30 m images and Landsat-5 Thematic Mapper(TM) images were used to classify based on maximum likelihood classification(MLC) algorithms. In order to validate the precision of the classifications, this study used aerial photography classification maps as training samples because of their high accuracy. The accuracy of the derived classes was assessed with the discrete multivariate technique called KAPPA accuracy. The results indicate:(1) training samples are important to classification results.(2) Image classification accuracy is always affected by areal fraction and aggregation degree as well as by diversities and patch shape.(3) The core zone area is protected better than buffer zone and experimental zone wetland. The experimental zone degrades fast because of irrational development by humans.
[34] 马红梅,王苗苗,刘勇.

多源遥感数据土地覆被空间尺度效应探讨

[J].遥感信息,2017,32(2):149-155.

URL      [本文引用: 1]     

[ Ma H M, Wang M M, Liu Y.

Spatial scale effects of land cover based on multi-source remote sensing data

[J]. Remote Sensing Information, 2017,32(2):149-155. ]

URL      [本文引用: 1]     

/