Satellite Image Classification Method Using the Dempster–Shafer Approach

In this paper the problem of the supervised classification of satellite images is considered. A new image classification method focused on application under conditions when the training sample is (in particular, considerably) contaminated is proposed. The method is based on using the Dempster–Shafer evidence theory and is applicable both for hyperspectral and multispectral satellite images. Problems of organizing the supervised classification process and content of its constitutive procedures are presented. The developed method has been implemented algorithmically and in software. Results obtained in the classification of hyperspectral images by the proposed method testify to its efficiency.

     

Identification of Forest Vegetation Using Airborne Hyperspectral Data

Izvestiya, Atmospheric and Oceanic Physics - Computational schemes for forest vegetation identification on airborne hyperspectral images are implemented using the feature of the red edge in the...     

基于无人机高光谱特征的红树林种群识别研究

红树林种群的组成和分布对于红树林生态系统的保护和恢复至关重要。本研究以漳江口红树林保护区为研究对象,通过获取无人机高光谱影像,进行光谱特征分析、光谱微分变换和包络线去除,提取了911组17个光谱特征参数,通过逐步判别分析筛选出13个用于决策树构建的特征参数,最终通过C5.0决策树模型获得了研究区红树林种群的分布状况。结果表明,漳江口红树林保护区植被种群呈现自上到下不同类型的分布情况,研究区上部以桐花树和秋茄混合类型为主,中间区域呈现白骨壤、桐花树和秋茄三者共生的现状,研究区下部则以白骨壤分布为主,伴生有少量的秋茄。通过混淆矩阵计算,得到研究区总体分类精度为 87.95%,Kappa系数为 83.81%,具有较好的精度。研究结果可为区域红树林湿地保护提供数据支撑,为红树林种群识别研究提供方法参考。     

基于空谱融合与AlexNet算法的滨海湿地植被分类研究

为提高遥感影像融合质量,提升资源一号(ZY-1 02D)高光谱遥感影像滨海湿地植被分类精度,提出将ZY-1 02D高光谱影像与空间分辨率为10 m的哨兵2号(Sentinel-2)影像进行Brovey融合,并通过搭建AlexNet卷积神经网络对ZY-1 02D高光谱影像和Brovey融合影像的滨海湿地植被进行分类,与支持向量机、随机森林和BP神经网络分类算法进行精度对比。研究结果表明:经Brovey融合后,AlexNet、支持向量机、随机森林和BP神经网络算法的植被分类总体精度分别提高15.60%、7.00%、14.80%和10.00%,Kappa系数提高了21.35%、9.93%、18.97%、12.85%;基于Brovey影像融合与AlexNet算法的植被分类精度最高,总体精度为92.40%,Kappa系数为89.42%。空谱融合配合AlexNet卷积神经网络有效解决了高光谱遥感影像在滨海湿地植被分类应用中精度较低的问题,为滨海湿地植被资源动态监测提供技术和方法支撑。     

Brown Carbon and Black Carbon in the Smoky Atmosphere during Boreal Forest Fires

We have investigated the variability of smoke aerosol absorbing ability with variations in the content of brown carbon (BrC) and black carbon (BC). Using monitoring data on radiative characteristics of smoke aerosol at AERONET stations and the spatial distribution of aerosol optical depth (AOD) obtained by the MODIS spectrometer (Terra satellite), we have detected large-scale smokes during boreal forest fires in Russia and Canada (1995–2012). The spatial distribution (50°–70° N, 95°–125° W) and temporal variability (at AERONET station Fort McMurray) of AOD during the smoking of a part of Canada in July 2012 have been analyzed. AOD probability distributions for July 14–18, 2012, and an estimate of aerosol radiative forcing of smoke aerosol at the upper boundary of the atmosphere have been obtained. We have proposed a technique for the diagnostics of BrC and BC in smoke aerosol particles from the spectral dependence of the imaginary part of the refractive index. At a wavelength of 440 nm, the contributions of BrC and BC to the smokeaerosol absorbing abitity can be comparable in magnitude. In many cases, the absorption spectra of smoke aerosol can be adequately approximated by either power or exponential functions. The presence of BrC in smoke-aerosol particles highly extends the variety of observed absorption spectra in a smoky atmosphere and spectral dependences of single scattering albedo. In the spectral range of 440–1020 nm, the radiative characteristics of smoke aerosol are largely contributed by its fine mode.     

The effect of atmospheric circulation on the evolution and radiative forcing of smoke aerosol over European Russia during the summer of 2010

The evolution of smoke plume over European Russia (ER) during the massive forest and peatbog fires of summer 2010 has been studied using observations of aerosol optical depth (AOD) from MODIS instruments (both Aqua and Terra platforms), objective analysis of meteorological fields performed at the Russian Hydrometeorological Research Center, NCEP/NCAR reanalysis, as well as upper air data. A relation between the structure inhomogeneities of the AOD field and regional atmospheric circulation has been found. It is shown that, on August 5–9, 2010, the maximum of smoke pollution did complete turn around Moscow, while remaining at a distance of 200 to 650 km from the megacity. Both regionally averaged shortwave aerosol radiative forcings (ARFs) at the top and the bottom of the atmosphere are estimated for the period of extreme smoke pollution over ER. The spatial distributions of ARF values over the territory of the region and the estimates of the local and spatially distributed thermal effects of smoke aerosol are given. It is shown that, on August 5–9, 2010, the spatial distribution of AOD and the calculated thermal effects of smoke aerosol were in agreement with the spatial distributions of air-temperature anomalies observed in the lower 1.5-km layer of the atmosphere. MODIS’s AOD data obtained during the wildfires were validated by AOD observations from the CIMEL sun photometer operated at the AERONET station Zvenigorod.     

Pattern Recognition of Forest Vegetation from Hyperspectral Airborne and Multichannel Satellite Data of High Spatial Resolution: Comparison of Results and Estimation of Their Accuracy

Izvestiya, Atmospheric and Oceanic Physics - The pattern recognition of a forest surface from remote sensing data (both airborne hyperspectral data and WorldView-2 multichannel satellite data of...     

Mapping urban heat islands of arctic cities using combined data on field measurements and satellite images based on the example of the city of Apatity (Murmansk Oblast)

This article presents the results of a study of the urban heat island (UHI) in the city of Apatity during winter that were obtained according to the data of field meteorological measurements and satellite images. Calculations of the surface layer temperature have been made based on the surface temperature data obtained from satellite images. The experimental data on air temperature were obtained as a result of expeditionary meteorological observations, and the experimental data on surface temperature were obtained based on the data of the space hyperspectral Moderate-Resolution Imaging Spectroradiometer (MODIS) system, channels 31 and 32 (10.78–11.28 and 11.77–12.27 micrometers, respectively). As a result of the analysis of temperature fields, an intensive heat island (up to 3.2°C) has been identified that was estimated based on the underlying surface temperature, and its mean intensity over the observation period significantly exceeds the representative data for European cities in winter. It has also been established that the air temperature calculated according to the MODIS data is systematically higher under winter conditions than the air temperature from direct measurement data.     

基于HICO波段的滨海土壤盐分遥感反演研究

本研究以黄河三角洲滨海盐渍土为例,尝试使用HICO （Hyperspectral Imager for the Coastal Ocean）高光谱影像结合现场实测高光谱数据进行表层土壤全盐含量的反演。采用波段组合的方法建立光谱参量,通过相关分析筛选出敏感光谱参量,以决定系数R2选出最佳模型;利用HICO影像反射率与实测高光谱反射率之间的关系,对模型进行修正,并应用于影像。研究发现,比值（RI）、差值（DI）波段组合方法建立的光谱参量与表层土壤全盐含量的相关性明显提高。DI_{（845,473）}、DI_{（839,490）}、DI_{（845,496）}及DI_{（839,501）}的幂函数模型效果最好,且验证决定系数R2均大于0.86,相对分析误差RPD>3,RMSE较小。此外,HICO遥感影像的模型反演结果较为一致,能够反映表层土壤全盐含量的分布。研究显示,利用高光谱数据进行表层土壤全盐含量的反演建模具有可行性,可为区域表层土壤全盐含量的定量反演提供参考。     

高光谱特征波段在海洋典型溢油油种识别中的精度评估

溢油种类精准识别对快速有效地治理污染具有重要的意义,高光谱遥感在海面溢油油种识别中至关重要。为探讨海洋典型溢油油种识别的高光谱特征波段范围,通过设计室外模拟溢油实验,在获取原油、燃料油、柴油、汽油和棕榈油等5种油种的实测高光谱数据基础上,运用因子分析和光谱标准偏差分析法遴选溢油油种的光谱特征波段,并利用支持向量机模型开展基于光谱特征波段的油种识别精度评价。结果表明,基于光谱标准偏差分析和因子分析获得的特征波段的油种识别精度分别是83.33%和90.74%,与基于全波段的识别精度相比,整体精度分别提高了3.7%和11.11%。选取的特征波段(360～540 nm, 560～600 nm, 610～630 nm, 640～660 nm)可作为5种油种相互区分的最佳高光谱波段。     

一种基于众数赋值的高光谱图像地物分类方法

提出了一种融合监督分类与非监督分类结果的高光谱遥感影像分类新方法——众数赋值分类法。采用ISODATA非监督分类方法对高光谱遥感影像进行分类,并对非监督分类结果的图斑进行标记,同时用最大似然法(ML)和支持向量机(SVM)法进行监督分类,然后以监督分类结果对非监督分类后各斑块进行类别赋值。方法是:统计每个非监督分类斑块中由监督分类所获得的各类别像元数及所占比例,将非监督分类斑块的类别赋予所占比例最高的监督分类结果的类别,最终获得高光谱图像分类结果。研究表明:(1)非监督分类类别数量大于10时,其与ML分类结果融合的总体分类精度和Kappa系数均较监督分类法的分类结果好;(2)ML和20个类别的ISODATA分类结果融合的总体精度最高,为87.35%,比单独ML的总体精度高约2个百分点;(3)SVM和10个类别的ISODATA分类结果融合的总体精度提高最大,较SVM的总体精度提高近3个百分点;(4)随着非监督分类类别数量的增多,分类结果的总体精度呈现由低到高再到低的变化过程。     

基于地物光谱可分性的CHRIS 高光谱影像波段选择及其分类应用

本文以黄河口湿地为研究区,应用覆盖该区域的CHRIS高光谱遥感影像,提出了一种基于地物光谱可分性的滨海湿地高光谱影像波段选择方法。该方法利用研究区的7种典型地物的110余条现场实测地物光谱曲线,通过分析比对地物两两之间的光谱可分度,确定地物类型之间的光谱可分区间,基于此选取CHRIS高光谱影像的地物分类特征波段,应用三种经典的监督分类方法(支持向量机法SVM、人工神经网络法ANN和光谱角制图法SAM)开展利用全波段的和利用本文方法选择特征波段的分类对比实验。结果表明:(1)基于光谱可分性特征波段的方法较全波段分类精度有所提高,其中ANN分类精度最高,为82.52%,较全波段分类精度提高了约为5.1%;(2)芦苇、水体、黄河水和裸滩4种地物的识别能力高,生产者精度都在80%以上;(3)碱蓬的用户精度提升最为明显,约在7%。     

Models of Pattern Recognition and Forest State Estimation Based on Hyperspectral Remote Sensing Data

Izvestiya, Atmospheric and Oceanic Physics - Model applications of airborne hyperspectral remote sensing data for the recognition of forest stand objects and parameterization of the environmental...     

Evaluation of the Species Composition and the Biological Productivity of Forests Based on Remote Sensing Data with High Spatial and Spectral Resolution

Izvestiya, Atmospheric and Oceanic Physics - The application of hyperspectral remote sensing of high spatial resolution is compared to conventional ground-based forest surveys on sample plots and...     

The radiative and thermal effects of smoke aerosol over the region of Moscow during the summer fires of 2010

The shortwave radiative forcings of smoke aerosol in the cloudless atmosphere during the summer fires of 2010 in European Russia were quantitatively estimated for the land surface and the atmospheric upper boundary from measurement data obtained at the Zvenigorod Scientific Station of the Obukhov Institute of Atmospheric Physics (OIAP ZSS), Russian Academy of Sciences. Variations in the temperature of the surface air layer due to the smoke-induced attenuation of incoming solar radiation were estimated. The most intensive smoke generation in the atmosphere was observed on August 7–9, 2010, when the maximum aerosol optical thickness amounted to more than 4.0 at a wavelength of 550 nm. In this case, the albedo of single aerosol scattering amounted to ∼0.95–0.96 and the asymmetry factor amounted to ∼0.69–0.70. The maximum shortwave radiative forcing of aerosol amounted to about −360 W/m² for the land surface and almost −150 W/m² for the atmospheric upper boundary. During the period of intensive smoke generation, the cooling of the atmospheric surface layer over daylight hours (12 h) amounted, on average, to ∼6°C. The power character of the dependence of the shortwave radiative forcing of aerosol for the land surface on aerosol optical thickness up to its values exceeding 4.0, which was revealed earlier on the basis of data on aerosol optical thickness (up to 1.5) obtained at the OIAP ZSS during the summer forest and peatbog fires of 2002 in the region of Moscow, was supported.     

海南红树林根系土壤中可培养细菌的多样性分析

为研究海南红树林根系土壤细菌多样性,充分发掘中国红树林特有微生物资源,作者于2012年12月份从10份采集于海南北港岛红树林根系淤泥样品中分离可培养海洋细菌,并基于16S rRNA基因序列分析的方法对样品中的细菌多样性进行研究。共分离得到122株海洋细菌,经并菌后选取其中的70株代表性菌株测序,结果发现它们分为4个类群:变形菌门(主要为γ-变形菌纲)(Proteobacteria)、拟杆菌门(Bacteroidetes)、厚壁菌门(Firmicutes)和放线菌门(Actinobacteria),分别占66%,19%,11%和4%。共代表30个属47个种,其中有9个潜在新种和2个潜在新属。本研究结果表明,北港岛红树林中存在着丰富的细菌多样性,很多类群可能是红树林特有细菌物种。     

协同主动学习和半监督方法的海冰图像分类

海冰遥感光谱影像分类中标签样本难以获取,导致海冰分类精度难以提高,但是大量包含丰富信息的未标签样本却没有得到充分利用,针对这种情况,提出一种协同主动学习和半监督学习方法用于海冰遥感图像分类。在主动学习部分,结合最优标号和次优标号、自组织映射神经网络以及增强的聚类多样性算法来选择兼具不确定性和差异性的样本参与训练;在半监督学习部分,利用直推式支持向量机,并且融合主动学习思想从大量未标签样本中选取相对可靠且包含一定信息量的样本进行迭代训练;然后协同主动学习分类结果和半监督分类结果,通过一致性验证保证所加入伪标签样本的正确性。为了验证方法的有效性,分别采用巴芬湾地区30 m分辨率的Hyperion高光谱数据(验证数据为15 m分辨率的Landsat-8数据)和辽东湾地区15 m分辨率的Landsat-8数据(验证数据为4.77 m分辨率的Google Earth数据)进行海冰分类实验。实验结果表明,相对其他传统方法,该协同分类方法可以在只有少量标签样本的情况下,充分利用大量未标签样本中包含的信息,实现快速收敛,并获得较高的分类精度(两个实验的总体精度分别为90.003%和93.288%),适用于海冰遥感图像分类。     

高光谱影像分类及组合分类器的应用

高光谱影像包含有丰富的地物类别信息,但由于受分类方法等因素的限制,这些信息并不能得到充分的挖掘。组合分类器由于可以实现各分类结果信息的互补,因而总有着较各单分类器效果更好的分类性能。将组合分类方法引入到高光谱影像的分类实践中,并对组合的关键问题一组合规则进行探讨,提出了混合组合规则的概念。实验表明,利用该组合分类方法对高光谱影像进行分类,可以得到良好的处理结果。     

基于无人机高光谱遥感技术对内陆养殖池塘水质监测的研究

为准确、快速、全面获取内陆养殖小区池塘及尾水处理池水体水质变化情况,建立养殖水环境实时预警调控及数字化管控机制,选择浙江湖州市集中连片的养殖小区为试验区,于2020年12月采用搭载GaiaSky-mini高光谱相机的无人机进行试验区近地遥感图像采集,并进行图像拼接、辐射校正和几何校正等预处理;然后对反射波段进行差值、比...     

基于深度置信网络(DBN)的赤潮高光谱遥感提取研究

赤潮是严重的海洋灾害,有效监测赤潮对于保护海洋生态环境具有重要意义。高光谱遥感具有光谱分辨率高、图谱合一等优势,适合于海洋赤潮监测。深度学习是机器学习领域的前沿,为高光谱遥感分类提供了新的思路。深度置信网络(Deep Belief Network,DBN)兼具监督分类与非监督分类的特点,通过构建DBN模型,将DBN应用于赤潮灾害遥感监测中,应用渤海机载高光谱遥感数据开展赤潮分类,以达到提取高光谱图像中赤潮水体范围的目的。通过设置对照实验,对比经典的SVM监督分类方法与ISODATA非监督分类方法,发现DBN模型在相同实验条件下具有更高的分类精度,赤潮遥感提取精度提高了3%～11%。