基于高光谱的湿地植被分类研究

光谱特征的选择对于湿地植被的识别精度和效率有直接的影响。本文以萨克拉门托-圣华金三角洲为研究区,基于Hy Map航空高光谱遥感影像数据,分析湿地植被的一阶微分和二阶微分光谱特征。在上述分析的基础上基于均值置信区间的波段选择法对一阶微分、二阶微分进行波段选择,根据获取的有效特征波段构建特征集,利用C5决策树分类算法产生规则集,并对实验区的湿地植被进行了分类研究。结果表明:湿地植被的一阶微分、二阶微分能够突出不同湿地植被光谱曲线在不同波段的增速不同,利用均值置信区间的波段选择法能够对特征波段起到降维效果,根据降维后的特征波段采用C5决策树分类算法,可以实现湿地植被在物种水平上的识别,并达到较好的分类精度。     

基于微分变换的高光谱马尾松和杉木识别

高光谱遥感能分辨出地物间微小反射光谱差异信息,可用于解决林种遥感分类光谱识别的难题。利用Hyperion高光谱遥感影像,结合地面实测林种样地,对安徽省黄山市五城镇林区的马尾松和杉木进行识别。通过对Hyperion影像进行一阶、二阶微分变换,优化组合487～559 nm和681～742 nm光谱范围中反射差异明显的波段,再结合支持向量机(support vector machine,SVM)模型进行林种间分类识别。基于Hyperion影像像元反射率及其一阶和二阶微分光谱的分类识别总体精度分别达到76. 50%,81. 42%和88. 52%,对应Kappa系数分别为0. 528 4,0. 625 7和0. 769 1。结果表明,基于二阶微分变换的高光谱数据,通过SVM模型,可有效提高马尾松和杉木的识别精度,为高光谱遥感针叶林种分类识别提供了一种技术途径。     

盐渍化土壤光谱特征分析与建模

为建立土壤盐渍化遥感监测模型,选取宁夏回族自治区平罗县典型土壤盐渍化发生区域作为研究区,以野外原位光谱测量数据和实验室内测得的土壤含盐量与p H值数据为基础,进行高光谱数据处理,分析不同盐渍化程度土壤的光谱特征;对实测土壤光谱反射率进行倒数、对数、均方根及其一阶微分等光谱变换,计算高光谱指数;与土壤样本含盐量进行相关性分析,筛选盐渍化土壤的光谱特征波段,利用多元线性回归分析建立土壤盐渍化监测模型。研究结果表明:以倒数一阶微分变换后的940 nm和1 094 nm波段作为特征波段构建的土壤盐渍化遥感监测模型最优。     

基于光谱特征的湿地植物种类识别

光谱特征的选择对于湿地植被的识别精度和效率有直接的影响作用。以美国舍曼(Sherman)岛水域为研究区,基于Hy Map航空高光谱遥感影像数据,分析湿地植被的一阶微分光谱和光谱吸收特征,利用逐步判别分析法筛选识别精度较好的光谱特征参数参与C4.5决策树分类。结果表明:4种湿地植被的一阶导数光谱特征差异较小,吸收特征差异性相对较大;基于一阶微分光谱特征和光谱吸收特征利用C4.5决策树进行分类,可以实现湿地植被在物种水平上的识别,并达到较好的分类精度。     

美国西部黄松叶面积指数与高光谱分辨率CASI数据的相关分析

本文介绍了美国俄罗冈州西部黄松的叶面积指数(LAI)与小型航空光谱制图成像仪(CASI)获取的高光谱分辨率数据进行的相关分析。在试验场地上使用LAI-2000植物冠层分析仪测得8个LAI值(0.87—2.72)。对CASI数据进行一阶和二阶微分处理,以减少土壤背景光谱对森林光谱的影响。使用逐步回归分析方法探索LAI与CASI数据的关系。由回归分析产生多元线性方程和相应的拟合度(GOF)及估计LAI的标准误(SE)。结果说明光谱微分技术能够提高LAI和CASI数据的相关性,因而可以改善LAI的估计精度。如,对于单通道LAI预测的最高GOF值是0.681,SE是0.345,而经一阶和二阶光谱微分处理后,GOF被分别提高到0.904和0.898,SE被分别降低到0.189和0.195.     

特征变量选择结合SVM的耕地土壤Hg含量高光谱反演

为探讨应用高光谱数据反演耕地土壤重金属汞(Hg)含量,对原始光谱进行10 nm重采样和SG平滑处理,用不同光谱变换数据与土壤重金属Hg含量进行相关性分析,采用IRIV、Random Frog和PCC提取光谱特征波段,分别建立SVM与GWO-SVM土壤Hg含量高光谱反演模型,获取Hg含量最优反演路径。研究表明,一阶微分变换光谱后土壤光谱特征更明显;上述特征提取方法在不同程度上减少光谱数据冗余,保留有效变量信息;经灰狼算法优化后支持向量机模型反演精度提高,IRIV结合GWO-SVM预测精度更高,其验证集R²为0.894,RMSE为0.082,MAE为0.016。研究成果可为类似土壤重金属含量的反演提供借鉴。     

基于高光谱遥感的四种典型道路光谱特征分析

利用高光谱遥感的方法,采用美国ASD公司生产的FieldSpec 4(背负式非成像高光谱仪),光谱范围覆盖350—2 500 nm,开展了对水泥混凝土道路、沥青道路、砖路和泥路这4种最为常见的典型道路的高光谱数据野外观测。利用光谱均值、反射率计算、一阶导数、二阶导数、倒数后对数变换等方法,对原始光谱数据进行处理,分析对比4种道路的光谱曲线,研究4种道路各自的光谱特征并找到能很好地区分这4种不同道路类型的最佳波段。旨在探索不同道路类型的光谱特征,为道路识别与提取提供重要依据。     

水稻叶面积指数的高光谱遥感估算模型

通过不同氮素营养水平的水稻田间试验 ,采用单变量线性与非线性拟合模型和逐步回归分析 ,用1 999年试验数据为训练样本 ,建立水稻LAI的高光谱遥感估算模型 ,用 2 0 0 0年试验数据作为测试样本数据 ,对其精度进行评价和验证。结果表明 ,高光谱变量与LAI之间的拟合分析中 ,蓝边内一阶微分的总和与红边内一阶微分的总和的比值和归一化差植被指数是最佳的变量     

光谱波段宽度对森林叶片叶绿素含量反演的影响分析

随着传感器技术的发展,高光谱数据光谱的波段宽度逐渐变窄,如何从海量的光谱数据中找到最优的光谱波段反演叶绿素含量,成为研究的难点问题。本文在测量华中农业大学狮子山6种主要树种的光谱数据和叶绿素含量的基础上,利用叶绿素指数(CI)和回归方法反演叶绿素含量,并分析了波段宽度对反演叶绿素含量结果的影响,结果发现波段宽度会影响到叶绿素反演的精度,当波段宽度为30nm时,叶绿素含量与"红边"区域(700nm-730nm)和近红外区域(770nm-800nm)叶绿素指标(CI)间的线性关系较好,决定系数可达到77.62%,均方根误差为10.6ug/cm2。     

耕层土壤有机质高光谱间接估测模型

针对光学遥感技术只能获取表层土壤光谱信息而无法直接估测耕层土壤有机质含量的问题,探索建立基于表层土壤高光谱信息的耕层土壤有机质间接估测模型。以山东省济南市章丘区采集的76个表层、耕层土壤样本数据为基础,首先分析原始光谱反射率的光谱特征;然后利用反射率的一阶微分、平方根的一阶微分和对数倒数的一阶微分等方法对原始光谱反射率进行变换,并根据极大相关性原则选取估测因子;最后根据表层土壤有机质含量与耕层有机质含量间的内在关系,建立耕层土壤有机质含量的间接估测模型。结果表明,以557、1 621、2 107和2 316 nm波段对数倒数的一阶微分变换值和864 nm波段反射率平方根倒数一阶微分的变换值为估测因子,使用二次函数关系模型对耕层土壤有机质含量间接估测的精度最高,其决定系数R~2为0.784,平均相对误差为10.7%。研究表明,利用表层土壤高光谱信息间接估测耕层有机质含量可行有效。     

Hyperspectral discrimination of tree species with different classifications using single- and multiple-endmember

Discrimination of tree species with different ages is performed in three classifications using hyperspectral data. The first classification is between Broadleaves and pines; the second classification is between Broadleaves, Corsican Pines, and Scots Pines, and the third classification is between six tree species including different ages of Corsican and Scots Pines. These three classifications are performed by having single- and multiple-endmember and considering five different spectral measure techniques (SMTs) in combination with reflectance spectra (ReflS), first and second derivative spectra. The result shows that using single-endmember, derivative spectra are not useful for a more challenging classification. This is further emphasized in multiple-endmember classification, where all SMTs perform better in ReflS rather than derivative in all classifications. Furthermore, using derivative spectra, discrimination accuracy become more dependent on the type of SMTs, especially in single-endmember. By employing multiple-endmember, the within-species variation is significantly reduced, thereby, the remaining challenge in discriminating tree species with different ages is only due to the between-species similarity. Overall, discrimination accuracies around 92.4, 76.8, and 71.5% are obtained using original reflectance and multiple-endmember for the first, second, and third classification, which is around 14.3, 17, and 8.3% higher than what were obtained in single-endmember classifications, respectively. Also, amongst the five SMTs, Euclidean distance (in both single- and multiple-endmember) and Jeffreys–Matusita distance (in single-endmember and derivative spectra) provided the highest discrimination accuracies in different classifications. Furthermore, when discrimination become more challenging from the first to second and third classification, the performance difference between different SMTs is increased from 1.4 to 3.8 and 7.3%, respectively. The study shows high potential of multiple-endmember to be employed in remote sensing applications in the future for improving tree species discrimination accuracy.     

不同传感器的模拟植被指数对水稻叶面积指数的估测精度和敏感性分析

本研究利用水稻冠层高光谱数据,模拟NOAA-AVHRR,Terra-MODIS和Landsat-TM的可见光波段反射率数据,计算各传感器的多种植被指数(NDVI,RVI,EVI,GNDVI,GRVI和Red-edge RVI),比较植被指数模型对水稻LAI的估测精度,分析不同植被指数对LAI变化的敏感性.相对于红波段植被指数,红边比值植被指数(Red-edge RVI)和绿波段指数GRVI与LAI有更好的线性相关关系,而GNDVI和LAI呈现更好的对数相关关系.MODIS的Red-edge RVI指数不仅模型拟合的精度最高,还有独立数据验证的估测精度也最高,而且它的验证精度较拟合精度下降幅度最小;其次是绿波段构建的GNDVI和GRVI植被指数的估测精度,再次是NDVI和EVI的估测精度,而RVI的估测精度最差.敏感性分析发现,13个植被指数对水稻LAI的估测能力都随着LAI的增加而下降,但归一化类植被指数和比值类植被指数对LAI变化反应的差异明显,归一化类植被指数在LAI较低时(LAI<1.5)对LAI变化的反应开始非常敏感,但迅速下降,而比值类植被指数在LAI较低时,明显小于归一化类植被指数,之后随着LAI的增大(LAI>1.5)比值类植被指数对LAI的变化敏感性,则明显高于归一化类植被指数.Red-edge RVI和绿波段指数GRVI和LAI不仅表现了很好的线性相关关系,而且在LAI大于2.9左右保持较高的敏感性.     

夏季太湖叶绿素a浓度的高光谱数据监测模型

本文依据2004年7月的实测数据构建了太湖夏季叶绿素a浓度的实测光谱数据估计模型,并使用2004年8月的数据对模型进行了验证。调查样点覆盖了太湖内的典型水域,水样数据由无锡太湖环境监测站采集。样点的光谱数据用ASDFieldSpec野外光谱仪获取,每个样点测量10次,测量结果被转换为遥感反射率。对不同的波段组合进行比较分析后,从可解释性出发,最终选择了归一化指数表达式作为最佳波段组合,所建立的模型为：Chla（μg/L）=EXP（2.478 ＋16.378＊N66）,其中,N66为（R696 -R661） /（R696 ＋R661）。模型的R^2为0.9051,显著性p〈0.0001。与其他模型相比,本文的模型比较稳健,用于估计8月的叶绿素a浓度具有较小的绝对误差。本文的工作同时表明,在太湖的夏季相邻月份,可以使用实测光谱数据模型进行水体叶绿素a浓度的估计。     

Mapping of stream microhabitats with high spatial resolution hyperspectral imagery

 Maximum likelihood supervised classifications with 1-m 128 band hyperspectral data accurately map in-stream habitats in the Lamar River, Wyoming with producer's accuracies of 91% for pools, 87% for glides, 76% for riffles, and 85% for eddy drop zones. Coarser resolution 5-m hyperspectral data and 1-m simulated multiband imagery yield lower accuracies that are unacceptable for inventory and analysis. Both high spatial resolution and hyperspectral coverage are therefore necessary to map microhabitats in the study area. In many instances, the high spatial resolution hyperspectral (HSRH) imagery appears to map the stream habitats with greater accuracy than our ground-based surveys, thus challenging classical approaches used for accuracy assessment in remote sensing. Received: 9 April 2001 / Accepted: 8 October 2001     

Multi-Feature Classification Approach for High Spatial Resolution Hyperspectral Images

High spatial resolution hyperspectral images not only contain abundant radiant and spectral information, but also display rich spatial information. In this paper, we propose a multi-feature high spatial resolution hyperspectral image classification approach based on the combination of spectral information and spatial information. Three features are derived from the original high spatial resolution hyperspectral image: the spectral features that are acquired from the auto subspace partition technique and the band index technique; the texture features that are obtained from GLCM analysis of the first principal component after principal component analysis is performed on the original image; and the spatial autocorrelation features that contain spatial band X and spatial band Y, with the grey level of spatial band X changing along columns and the grey level of spatial band Y changing along rows. The three features are subsequently combined together in Support Vector Machine to classify the high spatial resolution hyperspectral image. The experiments with a high spatial resolution hyperspectral image prove that the proposed multi-feature classification approach significantly increases classification accuracies.     

Employing ground and satellite-based QuickBird data and random forest to discriminate five tree species in a Southern African Woodland

With the emergence of very high spatial and spectral resolution data set, the resolution gap that existed between remote-sensing data set and aerial photographs has decreased. The decrease in resolution gap has allowed accurate discrimination of different tree species. In this study, discrimination of indigenous tree species (n?=?5) was carried out using ground based hyperspectral data resampled to QuickBird bands and the actual QuickBird imagery for the area around Palapye, Botswana. The purpose of the study was to compare the accuracies of resampled hyperspectral data (resampled to QuickBird sensors) with the actual image (QuickBird image) in discriminating between the indigenous tree species. We performed Random Forest (RF) using canopy reflectance taking from ground-based hyperspectral sensor and the reflectance delineated regions of the tree species. The overall accuracies for classifying the five tree species was 79.86 and 88.78% for both the resampled and actual image, respectively. We observed that resampled data set can be upscale to actual image with the same or even greater level of accuracy. We therefore conclude that high spectral and spatial resolution data set has substantial potential for tree species discrimination in savannah environments.     

Multispectral sensor spectral resolution simulations for generation of hyperspectral vegetation indices from Hyperion data

The use of multispectral satellite sensors for generation of hyperspectral indices is restricted because of their coarse spectral resolutions. In this study, we attempted to synthesize a few of these hyperspectral indices, viz. RedEdge Normalized Difference Vegetation Index (NDVI₇₀₅), Plant Senescence Reflectance Index (PSRI) and Normalized-Difference-Infrared-Index (NDII), for crop stress monitoring at regional scale using multispectral images, simulated from Hyperion data. The Hyperion data were resampled and simulated to corresponding spatial and spectral resolutions of AWiFS, OCM-2 and MODIS sensors using their respective filter function. Different possible combinations of two bands (i.e. simple difference, simple ratio and normalized difference) were computed using synthetic spectral bands of each sensor, and were regressed with NDVI₇₀₅, PSRI and NDII. Models with highest correlation were selected and inverted on Hyperion data of another date to synthesize respective multispectral indices. Synthetic broad band indices of multispectral sensors with their respective narrow band indices of Hyperion were found to be in good agreement.     

The Extraction of Urban Surface Water from Hyperspectral Data Based on Spectral Indices

To prevent confusion between water and buildings in the extraction of urban surface water from hyperspectral data, we analyzed the spectra of shadows and water in hyperspectral images, and proposed an anti-shadow water extraction method. This method first uses the normalized difference vegetation index (NDVI) for initial water extraction, then uses the height of the reflectance peak at 588 nm to eliminate the shadow of buildings. The method was validated by two hyperspectral datacubes, which were obtained for Jiaxing City and Zhoushan City in Zhejiang Province, China. Compared to the common spectral indices used to extract a water body, such as the NDVI, normalized difference water index, hyperspectral difference water index, and index of water index, the proposed method could effectively eliminate the shadow of buildings. The commission error reduced from more than 40% to about 15%, and the Kappa coefficient was increased from 60 and 70% to over 80% for the two datacubes. This indicated that the proposed method can inhibit the shadow of buildings and does not have a regional dependence.     

基于高光谱曲线峰度、偏度的植被叶绿素含量反演模型研究

本文分析了高光谱反射率及红边位置与叶片绿度的相关性,建立了基于敏感波段和红边位置的叶绿素估算模型。通过对不同叶绿素含量高光谱曲线特征的分析,提出了基于高光谱曲线峰度和偏度的叶绿素估算新思路,并分别建立基于原始光谱560-760nm波段和一阶导数光谱660-760nm波段对应峰度、偏度的叶绿素反演模型。结果表明,法国梧桐、无花果和白毛杨基于敏感波段的叶绿素含量反演模型的拟合度,与传统估算模型相比,本文提出的新估算模型可以明显提高高光谱反演叶绿素含量的能力。     

利用无人机高光谱影像的红树林群落物种分类

近年来红树林群落中物种结构简单、功能退化等环境问题日趋严重,为了及时准确掌握红树林群落的物种空间格局与分布,本文首先基于深圳福田红树林自然保护区无人机高光谱影像,利用归一化差值植被指数和归一化潮间红树林指数提取植被区域;然后在植被区域根据最佳指数法选取信息量大、波段相关性小的波段组合,分别采用基于像素支持向量机分类方法和面向对象影像分类方法对红树林物种进行分类。试验结果表明,基于像素支持向量机分类方法的总体精度为81.03%;利用面向对象影像分类方法的总体精度为85.58%。面向对象影像分类方法能有效去除椒盐噪声,充分利用对象光谱、形状及纹理信息,提供更准确的红树林分布信息。