融合Landsat ETM+和MODIS数据估算高时空分辨率地表短波反照率

提出一种通过融合高空间低时间分辨率、低空间高时间分辨率地表短波反照率,来估算高时空分辨率地表短波反照率的方法。首先,利用Landsat ETM+数据,通过窄波段到宽波段的转换得到一景或多景空间分辨率较高的ETM+蓝天空短波反照率;然后,在MODIS短波反照率产品基础上,以天空光比例因子为权重,得到空间分辨率较低的MODIS蓝天空短波反照率;最后,利用STARFM(Spatial and Temporal Adaptive Reflectance Fusion Model)模型融合ETM+短波反照率的空间变化信息和MODIS短波反照率的时间变化信息,得到高时空分辨率的地表短波反照率。针对STARFM模型在异质性区域估算精度降低的问题,通过以MODIS反照率影像各像元的端元(各地类)反照率取代MODIS像元反照率来提取时空变化等信息参与STARFM模型的融合过程,达到提高异质性区域估算精度的目的。结果显示,直接利用STARFM模型估算得到的高空间分辨率地表短波反照率处在合理的精度范围内(RMSE0.02),用改进后的STARFM模型估算得到的异质性区域短波反照率和真实ETM+短波反照率间的相关系数增大。     

Landsat8和MODIS融合构建高时空分辨率数据识别秋粮作物

本文利用Wu等人提出的遥感数据时空融合方法 STDFA(Spatial Temporal Data Fusion Approach)以Landsat 8和MODIS为数据源构建高时间、空间分辨率的遥感影像数据。以此为基础,构建15种30 m分辨率分类数据集,然后利用支持向量机SVM(Support Vector Machine)进行秋粮作物识别,验证不同维度分类数据集进行秋粮作物识别的适用性。实验结果显示,不同分类数据集的秋粮作物分类结果均达到了较高的识别精度。综合各项精度指标分析,Red+Phenology数据组合对秋粮识别效果最好,水稻识别的制图精度和用户精度分别达到91.76%和82.49%,玉米识别的制图精度和用户精度分别达到85.80%和74.97%,水稻和玉米识别的总体精度达到86.90%。     

基于MODIS影像的超分辨率重建

针对4幅多时相250 m分辨率的MODIS影像,利用空间域中的超分辨率重建模型,进行了影像的匹配和运动参数估计,重建出采样率提高2倍的高分辨率影像。     

融合高时空分辨率数据估算植被净初级生产力

植被净初级生产力NPP(Net Primary Production)遥感估算与分析,有赖于高时空分辨率的遥感数据,但目前中高分辨率的遥感数据受卫星回访周期及天气的影响,在中国南方地区难以获取连续时间序列的数据,从而影响了高精度的区域植被净初级生产力的遥感估算。为此,提出一种基于多源遥感数据时空融合技术与CASA模型估算高时空分辨率NPP的方法。首先,利用多源遥感数据,即Landsat8 OLI数据与MODIS13Q1数据,采用遥感数据时空融合方法,获得了时间序列的Landsat8 OLI融合数据;然后,基于Landsat8 OLI时空融合数据,并采用CASA模型,以长株潭城市群核心区为例,进行区域植被NPP的遥感估算。研究结果表明,基于时间序列Landsat融合数据估算的30m分辨率的NPP具有良好的空间细节信息,且估算值与实测值的相关系数达0.825,与实测NPP数据保持了较好的一致性。     

主题学习和稀疏表示的MODIS图像超分辨率重建

针对MODIS图像分辨率受传感器限制和噪声干扰,且分辨率局限在一定水平等问题,提出一种采用主题学习和稀疏表示的MODIS图像超分辨率重建方法,该方法通过双边滤波将MODIS图像的平滑及纹理部分分离,并将纹理部分看成是由若干"文档"组成的训练样本;运用概率潜在语义分析提取"文档"的潜在语义特征,从而确定"文档"所属的"主题"。在此基础上,针对每个主题所对应的图像块,采用改进的K-SVD方法训练若干适用于不同主题的高低分辨率字典对,从而可以运用这些字典对,通过稀疏编码实现测试图像相应主题块的超分辨率重建。实验结果表明,重建图像在视觉效果和PSNR等指标上均优于传统方法。     

一种自适应正则MAP超分辨率重建方法

提出了一种简单、通用的基于自适应正则化技术的MAP超分辨率重建新方法,该方法可以自适应地选择正则化参数,并充分利用中间重建结果的信息,不断对其进行更新,通过迭代过程得到最终的重建影像。利用模拟序列影像和真实序列影像分别对本文方法进行了验证,结果表明,本文方法可以根据不同影像序列的特点,自适应确定相应的正则化参数,并能找到最优解,适用性较强。     

MODIS和HJ-1 CCD数据时空融合重构NDVI时间序列

遥感数据反演高时空分辨率NDVI对监测植被动态变化过程具有重要意义,然而受天气影响,单颗卫星难以提供时间连续的高空间分辨率NDVI数据。以华北平原中东部为实验区,联合HJ-1 CCD数据和MODIS数据,对STARFM算法进行了改进,(1)考虑了不同地物对光谱响应的差异,为减少分类错误利用统计学上()对分类数据进行筛选,按照不同地物类型分别利用线性拟合方法修改光谱距离权重;(2)定义了预测半径,对HJ-1 CCD数据因外界影响而缺失的影像进行了预测。结果表明,与真实影像相比,预测结果呈现了较好的空间一致性,相关系数均达到了极显著相关,改进算法的预测精度要高于原算法。利用该方法将HJ-1 CCD NDVI的空间变化信息与MODIS NDVI时间变化信息有机结合重构了高时空分辨率NDVI序列,有效补充了HJ-1CCD NDVI的缺失数据集。     

基于压缩感知的航空影像超分辨率重建

详细介绍了压缩感知的研究现状和发展趋势,对压缩感知理论的原理、稀疏基的选取、测量矩阵的构造和信号的重建进行了详细阐述。在图像超分辨率重建技术的基本模型的基础上,基于压缩感知的图像超分辨率重建模型,使用小波基作为稀疏基,并使用改进的正则化正交匹配追踪算法对单幅图像进行超分辨率重建。最后,进行仿真实验,实现了基于压缩感知的单幅图像超分辨率重建,并且和传统的超分辨率重建算法进行对比。结果表明,基于压缩感知的图像超分辨率重建算法,取得了比较好的效果。     

利用地面靶标的超分辨率重建图像分辨率评价

针对超分辨率重建后图像分辨率指标计算困难的问题,该文提出一种利用地面靶标的重建图像分辨率计算方法。依据靶标原理和分辨率测量需求设计出改进的线阵和辐射状地面靶标;实地拍摄不同分辨率的靶标图像,建立真实分辨率图像库;采用卷积神经网络方法对拍摄图像进行超分辨率重建,并利用靶标计算重建后图像的分辨率,并与MSE、RMSE、PSNR等指标进行对比,进一步验证本文方法的可行性。通过实验可以得出,线阵与辐射状靶标对图像分辨率的评价结果与主观评价结果一致。实验证明,该方法能够计算出超分辨率重建后图像分辨率值,研究成果具有实际应用价值。     

基于导向滤波与迭代反向投影的遥感影像超分辨率重建

针对单帧遥感影像采取迭代反投影方法进行超分辨率重建时,重建图像的强边缘存在锯齿效应,在分析了导向滤波算法原理后,本文提出了一种将上述算法引入迭代过程来处理图像误差的方法,以进一步提升图像的高频信息,提高图像的重建质量。选取同时间不同地物的遥感影像作为实验数据,实验结果表明,本文重建的结果与双三次插值方法、边缘导向插值方法和迭代反投影方法相比,在客观评价指标上均有提高,改善了重建影像的纹理细节。本文提出的超分辨率重建方法,可以使重建影像提供更多的高频信息,具有较好的稳定性和鲁棒性。     

Improving the mapping of crop types in the Midwestern U.S. by fusing Landsat and MODIS satellite data

Mapping crop types is of great importance for assessing agricultural production, land-use patterns, and the environmental effects of agriculture. Indeed, both radiometric and spatial resolution of Landsat’s sensors images are optimized for cropland monitoring. However, accurate mapping of crop types requires frequent cloud-free images during the growing season, which are often not available, and this raises the question of whether Landsat data can be combined with data from other satellites. Here, our goal is to evaluate to what degree fusing Landsat with MODIS Nadir Bidirectional Reflectance Distribution Function (BRDF)-Adjusted Reflectance (NBAR) data can improve crop-type classification. Choosing either one or two images from all cloud-free Landsat observations available for the Arlington Agricultural Research Station area in Wisconsin from 2010 to 2014, we generated 87 combinations of images, and used each combination as input into the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) algorithm to predict Landsat-like images at the nominal dates of each 8-day MODIS NBAR product. Both the original Landsat and STARFM-predicted images were then classified with a support vector machine (SVM), and we compared the classification errors of three scenarios: 1) classifying the one or two original Landsat images of each combination only, 2) classifying the one or two original Landsat images plus all STARFM-predicted images, and 3) classifying the one or two original Landsat images together with STARFM-predicted images for key dates. Our results indicated that using two Landsat images as the input of STARFM did not significantly improve the STARFM predictions compared to using only one, and predictions using Landsat images between July and August as input were most accurate. Including all STARFM-predicted images together with the Landsat images significantly increased average classification error by 4% points (from 21% to 25%) compared to using only Landsat images. However, incorporating only STARFM-predicted images for key dates decreased average classification error by 2% points (from 21% to 19%) compared to using only Landsat images. In particular, if only a single Landsat image was available, adding STARFM predictions for key dates significantly decreased the average classification error by 4 percentage points from 30% to 26% (p < 0.05). We conclude that adding STARFM-predicted images can be effective for improving crop-type classification when only limited Landsat observations are available, but carefully selecting images from a full set of STARFM predictions is crucial. We developed an approach to identify the optimal subsets of all STARFM predictions, which gives an alternative method of feature selection for future research.     

Automatic and improved radiometric correction of Landsat imagery using reference values from MODIS surface reflectance images

Radiometric correction is a prerequisite for generating high-quality scientific data, making it possible to discriminate between product artefacts and real changes in Earth processes as well as accurately produce land cover maps and detect changes. This work contributes to the automatic generation of surface reflectance products for Landsat satellite series. Surface reflectances are generated by a new approach developed from a previous simplified radiometric (atmospheric + topographic) correction model. The proposed model keeps the core of the old model (incidence angles and cast-shadows through a digital elevation model [DEM], Earth–Sun distance, etc.) and adds new characteristics to enhance and automatize ground reflectance retrieval. The new model includes the following new features: (1) A fitting model based on reference values from pseudoinvariant areas that have been automatically extracted from existing reflectance products (Terra MODIS MOD09GA) that were selected also automatically by applying quality criteria that include a geostatistical pattern model. This guarantees the consistency of the internal and external series, making it unnecessary to provide extra atmospheric data for the acquisition date and time, dark objects or dense vegetation. (2) A spatial model for atmospheric optical depth that uses detailed DEM and MODTRAN simulations. (3) It is designed so that large time-series of images can be processed automatically to produce consistent Landsat surface reflectance time-series. (4) The approach can handle most images, acquired now or in the past, regardless of the processing system, with the exception of those with extremely high cloud coverage. The new methodology has been successfully applied to a series of near 300 images of the same area including MSS, TM and ETM+ imagery as well as to different formats and processing systems (LPGS and NLAPS from the USGS; CEOS from ESA) for different degrees of cloud coverage (up to 60%) and SLC-off. Reflectance products have been validated with some example applications: time series robustness (for a pixel in a pseudoinvariant area, deviations are only 1.04% on average along the series), spectral signatures generation (visually coherent with the MODIS ones, but more similar between dates), and classification (up to 4 percent points better than those obtained with the original manual method or the CDR products). In conclusion, this new approach, that could also be applied to other sensors with similar band configurations, offers a fully automatic and reasonably good procedure for the new era of long time-series of spatially detailed global remote sensing data.     

Fusing MODIS with Landsat 8 data to downscale weekly normalized difference vegetation index estimates for central Great Basin rangelands,USA

Data fused from distinct but complementary satellite sensors mitigate tradeoffs that researchers make when selecting between spatial and temporal resolutions of remotely sensed data. We integrated data from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor aboard the Terra satellite and the Operational Land Imager sensor aboard the Landsat 8 satellite into four regression-tree models and applied those data to a mapping application. This application produced downscaled maps that utilize the 30-m spatial resolution of Landsat in conjunction with daily acquisitions of MODIS normalized difference vegetation index (NDVI) that are composited and temporally smoothed. We produced four weekly, atmospherically corrected, and nearly cloud-free, downscaled 30-m synthetic MODIS NDVI predictions (maps) built from these models. Model results were strong with R² values ranging from 0.74 to 0.85. The correlation coefficients (r ≥ 0.89) were strong for all predictions when compared to corresponding original MODIS NDVI data. Downscaled products incorporated into independently developed sagebrush ecosystem models yielded mixed results. The visual quality of the downscaled 30-m synthetic MODIS NDVI predictions were remarkable when compared to the original 250-m MODIS NDVI. These 30-m maps improve knowledge of dynamic rangeland seasonal processes in the central Great Basin, United States, and provide land managers improved resource maps.     

自相似性和边缘保持分解的超分辨率重建算法

针对目前遥感图像超分辨率重建中存在边缘细节信息重建效果不佳的问题,本文提出了—种自相似性特征和边缘特征保持分解的超分辨率重建方法。首先,为了充分利用原始低分辨率图像自身的相似性信息,通过局部自相似性重建方法得到图像的初始重建结果;然后,为进一步增加不同尺度的边缘信息,采用加权最小二乘法对初始重建结果进行多尺度边缘保持分解,并对分解的细节层进行加权线性组合;最后,通过优化计算,得到融合多尺度边缘、细节信息及局部相似性特征的超分辨率重建图像。利用多组仿真和遥感卫星图像进行对比试验。结果表明,该方法可有效提升遥感图像的边缘信息和细节信息。     

一种拓展的半物理时空融合算法及其初步应用

Landsat 5卫星较低的时间分辨率(16天)使得其很难获得大区域的、时相一致的清晰影像数据集。本文发展了一种基于半物理模型的时空融合算法-即乘性调制融合算法,并借助多时序的MODIS反射率数据来生成多时相的Landsat TM/ETM+反射率合成影像,经镶嵌后得到区域尺度的高时空分辨率地表反射率数据集(Landsat TM/ETM+)。本文利用吉林省2006年—2011年的Landsat 5 TM地表反射率数据以及500 m的MOD09A1反射率产品来生成3个时相的Landsat 5 TM反射率合成数据,从而获得研究区在上述时相下地表反射率数据的镶嵌图。初步分析表明,所生成的Landsat 5 TM反射率数据的光谱分布特征与MOD09A1反射率数据较为一致,且图像在整体上光谱特征的连续性较好。     

一种基于 Contourlet 变换的遥感影像超分辨率重建改进方法

针对原有方法轮廓波(Contourlet)分解高频信息处理的不足,提出用零填充重采样代替双三次插值用于高频信息处理的新方法,该方法可使Contourlet分解后的高频信息重采样过程不引入噪声能量的同时获得理想的插值结果。改进方法利用Contourlet多分辨率分析的特征,以及零填充重采样方法较传统插值方法优越的性能,此两者相结合可改进原有方法的不足,提高单帧影像的超分辨率重建效果。经理论分析与试验验证,改进后方法的超分辨率重建结果无论在定性分析与定量对比上都要优于传统方法。     

Evaluating land cover changes in Eastern and Southern Africa from 2000 to 2010 using validated Landsat and MODIS data

In this study, we assessed land cover land use (LCLU) changes and their potential environmental drivers (i.e., precipitation, temperature) in five countries in Eastern & Southern (E&S) Africa (Rwanda, Botswana, Tanzania, Malawi and Namibia) between 2000 and 2010. Landsat-derived LCLU products developed by the Regional Centre for Mapping of Resources for Development (RCMRD) through the SERVIR (Spanish for “to serve”) program, a joint initiative of NASA and USAID, and NASA’s Moderate Resolution Imaging Spectroradiometer (MODIS) data were used to evaluate and quantify the LCLU changes in these five countries. Given that the original development of the MODIS land cover type standard products included limited training sites in Africa, we performed a two-level verification/validation of the MODIS land cover product in these five countries. Precipitation data from CHIRPS dataset were used to evaluate and quantify the precipitation changes in these countries and see if it was a significant driver behind some of these LCLU changes. MODIS Land Surface Temperature (LST) data were also used to see if temperature was a main driver too.Our validation analysis revealed that the overall accuracies of the regional MODIS LCLU product for this African region alone were lower than that of the global MODIS LCLU product overall accuracy (63–66% vs. 75%). However, for countries with uniform or homogenous land cover, the overall accuracy was much higher than the global accuracy and as high as 87% and 78% for Botswana and Namibia, respectively. In addition, the wetland and grassland classes had the highest user’s accuracies in most of the countries (89%–99%), which are the ones with the highest number of MODIS land cover classification algorithm training sites.Our LCLU change analysis revealed that Botswana’s most significant changes were the net reforestation, net grass loss and net wetland expansion. For Rwanda, although there have been significant forest, grass and crop expansions in some areas, there also have been significant forest, grass and crop loss in other areas that resulted in very minimal net changes. As for Tanzania, its most significant changes were the net deforestation and net crop expansion. Malawi’s most significant changes were the net deforestation, net crop expansion, net grass expansion and net wetland loss. Finally, Namibia’s most significant changes were the net deforestation and net grass expansion.The only noticeable environmental driver was in Malawi, which had a significant net wetland loss and could be due to the fact that it was the only country that had a reduction in total precipitation between the periods when the LCLU maps were developed. Not only that, but Malawi also happened to have a slight increase in temperature, which would cause more evaporation and net decrease in wetlands if the precipitation didn’t increase as was the case in that country. In addition, within our studied countries, forestland expansion and loss as well as crop expansion and loss were happening in the same country almost equally in some cases. All of that implies that non-environmental factors, such as socioeconomics and governmental policies, could have been the main drivers of these LCLU changes in many of these countries in E&S Africa. It will be important to further study in the future the detailed effects of such drivers on these LCLU changes in this part of the world.