Testing two methods for mapping water hyacinth (Eichhornia crassipes) in the Greater Letaba river system,South Africa: discrimination and mapping potential of the polar-orbiting Sentinel-2 MSI and Landsat 8 OLI sensors

ABSTRACT

Early detection and mapping of the spatio-temporal distribution of invasive water hyacinth (Eichhornia crassipes) in inland hydrological systems are vital for a number of water resource management-related reasons. Field surveys and current climate change projections (associated with longer dry spells, and shortened rain seasons) have shown that climate change and the rapid spread of aquatic invasive species are increasingly affecting inland surface water availability in semi-arid regions of Southern Africa. It is upon this premise that accurate, reliable, and timely information on the spatio-temporal distribution and configuration of water hyacinth is required in tracing their evolution and propagation in affected areas as well as in potential vulnerable areas. This work, therefore, attempts to test two robust push-broom multispectral sensors: Landsat 8 Operational Land Imager (OLI) and Sentinel-2 MultiSpectral Instrument (MSI) in identifying, detecting, and mapping the spatial distribution and configuration of invasive water hyacinth in a river system. The results of the study show that water hyacinth in small reservoirs can be mapped with an overall accuracy of 68.44% and 77.56% using Landsat 8 and Sentinel-2 data, respectively. The results further demonstrated the blue, red, red edge (RE) 1, short-wavelength infrared (SWIR)-1, and SWIR-2 of both satellite data sets as the critical and outstanding spectral regions in detecting and mapping water hyacinth from other land-cover types. Overall, the study highlights the unexploited prospects of the new noncommercial multispectral sensors in monitoring invasive species infestation from inland surface waterbodies in semi-arid regions (i.e. smaller reservoirs).     

Evaluation of Landsat 8 OLI imagery for unsupervised inland water extraction

The successful launch of the Landsat 8 satellite continues the Earth observation of the Landsat series, which has been taking place for nearly 40 years. With the increase in the band number and the improved spectral range compared with the previous Landsat imagery, it will be possible to expand the application of the new Landsat 8 imagery. The purpose of this study is to explore water extraction based on the new Landsat 8 Operational Land Imager (OLI) imagery. According to the specific inland water conditions (clear water, turbid water, and eutrophic water), a number of highly adaptable water indices are assessed for water extraction using Landsat OLI imagery. The results show that clear water is the easiest to extract among the different types of waterbodies, with the highest average accuracy of 97%. The highest-accuracy methods are the automated water extraction index for shadow pixels (AWEI_sh), the normalized difference water index using bands 4 and 7 (NDWI₄₇), and the normalized difference water index using bands 3 and 7 (NDWI₃₇), with accuracies of 98.55%, 95.50%, and 96.61%, corresponding to clear water, turbid water, and eutrophic water, respectively. Through the analysis of the different methods for optimal band selection, the seventh band OLI7 (shortwave infrared 2, SWIR-2) of Landsat OLI shows the best performance in water identification. When applying the water indices to water extraction, Otsu’s algorithm has been used to automatically select the water threshold. Using extensive experiments with Otsu’s algorithm and a manual method, it was found that Otsu’s algorithm can replace manual selection and has the ability to select an accurate threshold for water extraction.     

Non-photosynthetic vegetation biomass estimation in semiarid Canadian mixed grasslands using ground hyperspectral data,Landsat 8 OLI,and Sentinel-2 images

ABSTRACT

Research on quantifying non-photosynthetic vegetation (NPV) with optical remote-sensing approaches has been focusing on optically distinguishing NPV from green vegetation and bare soil. With a very similar spectral response curve to NPV, dry moss is a significant component in semiarid mixed grasslands and plays a large role in NPV estimation. However, limited attention has been paid to this role. We investigated the potential of optical remote sensing to distinguish NPV biomass in semiarid grasslands characterized by NPV, biological soil crust dominated by moss and lichen, and bare soil. First, hyperspectral spectral indices were examined to determine the most useful spectral wavelength regions for NPV biomass estimation. Second, multispectral red-edge indices and shortwave infrared (SWIR) indices were simulated based on Landsat 8 Operational Land Imager (OLI) and Sentinel-2A MultiSpectral Instrument band reflectance, respectively, to determine the most suitable multispectral indices for NPV estimation. The potential multispectral indices were then applied to Landsat 8 OLI images and Sentinel-2A images acquired in early, middle, peak, and early senescence growing seasons to investigate the potential of satellite images for quantifying NPV biomass. Our results indicated that hyperspectral red-edge indices, modified simple ratio, modified red-edge normalized difference vegetation index (mNDVI₇₀₅), and normalized difference vegetation index (NDVI₇₀₅) are better than the SWIR hyperspectral indices, including cellulose absorption index for quantifying NPV biomass. The simulated multispectral red-edge spectral indices (NDVI_red-edge and mNDVI_red-edge) demonstrate good and comparable performance on quantifying NPV biomass with SWIR multispectral indices (normalized difference index [NDI5 and NDI7] and soil-adjusted corn residue index). Nevertheless, the multispectral indices derived from Landsat 8 OLI and Sentinel-2 images have limited potential for NPV biomass estimation.     

Comparing the spectral signal of burned surfaces between Landsat 7 ETM+ and Landsat 8 OLI sensors

There is a long history of the use of Landsat data in burned land mapping mainly due to certain characteristics of the Landsat imagery including the spatial, spectral, and temporal data resolution, the low cost (Landsat data are now freely available), and the existence of an almost 35-year historical archive (excluding Landsat 1–3). Landsat 8 (Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS)) was launched on 11 February 2013 and it captures data in three new bands along with two additional thermal bands. However, is the spectral signal of burned surfaces in satellite remote-sensing data of Landsat series consistent and robust enough to allow the successful application of the techniques developed so far for Landsat 8? In this article, we compare the spectral signal of burned surfaces between Landsat 7 Enhanced Thematic Mapper Plus (ETM+) and Landsat 8 OLI sensors using five case studies that correspond to five large fire events in different biophysical environments in Greece, for which both Landsat 7 ETM+ and Landsat 8 OLI data were available. From the comparative analysis using histogram data plots of burned (post-fire image) and vegetated (pre-fire image) areas, spectral signature plots and separability indices of certain land-cover types, estimated using the same sampling areas over both satellite images, a general consistency was observed between the two sensors. Slight differences between the sensors were attributed to differences in the acquisition dates and were related to the type of vegetation rather than the sensors used to record the satellite images. Neither sensor provided improved discrimination over the other.     

基于局部空间自相关分析的时空数据融合

由于受到16d重访周期与云等对数据质量的影响,具有时间与空间连续性的Landsat 8OLI观测数据难以直接获取。考虑地物分布的空间自相关性,提出一种基于STARFM模型改进的局部自相关时空数据融合模型(LASTARFM),以新疆维吾尔族自治区喀什地区叶城县为研究区,利用Landsat 8OLI数据和MODIS数据的红光波段和近红外波段进行融合方法测试。结果表明:利用LASTARFM模型得到的融合影像,与真实影像NDVI相关系数达到0.92;在局部空间自相关性低的区域比STARFM模型影像反映出更多地物细节,具有更高的融合精度;在土地利用类型发生显著变化的区域与真实影像具有一定差异。     

Mapping concentrations of surface water quality parameters using a novel remote sensing and artificial intelligence framework

The deterioration of surface water quality occurs due to the presence of various types of pollutants generated from human, agricultural, and industrial activities. Thus, mapping concentrations of different surface water quality parameters (SWQPs), such as turbidity, total suspended solids (TSS), chemical oxygen demand (COD), biological oxygen demand (BOD), and dissolved oxygen (DO), is indeed critical for providing the appropriate treatment to the affected waterbodies. Traditionally, concentrations of SWQPs have been measured through intensive field work. Additionally, quite a lot of studies have attempted to retrieve concentrations of SWQPs from satellite images using regression-based methods. However, the relationship between SWQPs and satellite data is complex to be modelled accurately by using regression-based methods. Therefore, our study attempts to develop an artificial intelligence modelling method for mapping concentrations of both optical and non-optical SWQPs. In this context, a remote-sensing framework based on the back-propagation neural network (BPNN) is developed for the first time to quantify concentrations of different SWQPs from the Landsat8 satellite imagery. Compared to other methods, such as Support Vector Machine, significant coefficients of determination (R²) between the Landsat8 surface reflectance and concentrations of SWQPs were obtained using the developed Landsat8-based-BPNN models. The resulting R² values were 0.991, 0.933, 0.937, 0.930, and 0.934 for turbidity, TSS, COD, BOD, and DO, respectively. Indeed, these findings indicate that the developed Landsat8-based-BPNN framework is capable of developing highly accurate models for retrieving concentrations of different SWQPs from the Landsat8 imagery.     

Analysing changes of the Poyang Lake water area using Sentinel-1 synthetic aperture radar imagery

Poyang Lake is the largest freshwater lake in China. Monitoring changes of its water area is essential for the conservation of important wetlands and ecological resources, and plays an important role for sustainable water use and management. Landsat and Moderate-Resolution Imaging Spectroradiometer (MODIS) sensor images are widely used for mapping waterbodies, because of their sensitivity for spectral reflectance of water. However, studies using Landsat images have limited their investigations of changes of Poyang Lake to dry season due to the impairment by cloud cover. Further limited by the rather long 16 day revisit cycle, most existing studies build on the vague assumption that Poyang Lake undergoes only relatively slow changes during this season. MODIS, in contrast, provides a very short revisit period, but has been proven not to be able to assess the water area of Poyang Lake accurately due to low spatial resolution. Therefore, the contribution of this study is to investigate recent Poyang Lake water area changes both during high- and low-water period with unforeseen temporal and spatial resolution using Sentinel-1 synthetic aperture radar (SAR) imagery. More specifically, we aim at investigating Poyang Lake’s recent water area changes in intra-month scales. During the observation period from October 2014 to March 2016, October 2014 was the month with the largest max/min water coverage ratio. Water coverage of winter in 2014 and 2015 was completely different, as a severe drought happened in 2014 and an unusual winter flood happened in 2015. Thus, this study demonstrates the potential of using Sentinel-1 SAR data to reveal intra-month variations, benefiting from the sensor’s regular observation capabilities independent of weather conditions. It is shown that Sentinel-1 SAR data, with rapid availability and free-of-charge distribution policy, as well as relatively high spatial and temporal resolutions, is becoming an indispensable data source for a detailed monitoring of important inland waterbodies and wetlands.     

A water coverage extraction approach to track inundation in the Saskatchewan River Delta,Canada

Tracking surface water coverage changes is a complicated task for many regions of the world. It is, however, essential to monitor the associated biological changes and bioproductivity. We present a methodology to track contemporary water coverage changes using optical remote sensing and use it to estimate historical summer water coverage in a large river delta. We used a geographical information system automated routine, based on the modified normalized difference water index, to extract the surface water coverage area (SWCA) from optical satellite data sets using the surface water extraction coverage area tool (SWECAT). It was applied to measure SWCA during drought and flood peaks in the Saskatchewan River Delta in Canada, from Landsat, SPOT and RapidEye images. Landsat results compared favourably with Canadian National Hydro Network (CNHN) GeoBase data, with deviations between SWCA classifications and the base CNHN GeoBase shapefile of ~2%. Difference levels between the extracted SWCA layer from Landsat and the higher resolution commercial satellites (SPOT and RapidEye) were also less than 2%. SWCA was tightly linked to discharge and level measurements from in-channel gauges (r² > 0.70). Using the SWCA versus discharge relationship for the gauge with the longest record, we show that peak summer SWCA has declined by half over the last century, from 13% of our study area to 6%, with likely implications for fish and wildlife production.     

Inland waterbody mapping: towards improving discrimination and extraction of inland surface water features

Surface waterbodies in arid and semi-arid environments are threatened by both natural and anthropogenic pressures. Mapping the distribution of surface waterbodies is crucial for managing their dwindling quantities and quality. In this study, a fast and reliable method of water extraction has been introduced. A remote-sensing index called the simple water index (SWI) was formulated to differentiate waterbodies from vegetation class automatically, and to differentiate waterbodies from shadows or built-up areas (water-like features). Its performance was compared with the automated water extraction index (AWEI) and the modified normalized difference water index (MNDWI) on Landsat 8 Operational Land Imager (OLI) image of South Africa. The robustness of the algorithm was tested on images in Madagascar and the Democratic Republic of Congo (DRC) with different biomes. The overall accuracies and kappa coefficient (κ) were used to compare the performance of each index. The McNemar test was performed to assess the significance of the output map and the validation data set. The SWI showed the highest overall accuracy of 91.9% (κ = 0.83), whereas the AWEI and MNDWI yielded overall accuracies of 83.8% (κ = 0.65) and 78.4% (κ = 0.53), respectively. The McNemar test showed that there was no significant difference between the SWI map (p = 0.248), whereas both AWEI and MNDWI maps were significantly different from the validation data set at p = 0.041 and p = 0.013, respectively. The SWI approach reduces the thresholding problem by 50% over the conventional MNDWI and AWEI. It is expected that the SWI will also be useful for the accurate quantification of waterbodies for large areas.     

Assessing the effectiveness of Google Earth images for spatial enhancement of RapidEye multi-spectral imagery

RapidEye satellite images with high spatial resolution, affordable prices and having Red-Edge band have high potential for time series issues, especially in vegetation studies. Despite these beneficial properties, RapidEye images with 5 m spatial resolution are not sufficiently useful for some applications. According to this problem, enhancing the spatial resolution of RapidEye images can significantly improve the results of the subsequent processes on these images. Fusion of high spatial resolution with high spectral resolution images is known as an effective way to enhance the quality of multispectral remotely sensed images. Unfortunately, the lack of panchromatic band with high spatial resolution has been faced the procedure of improving the spatial resolution of RapidEye images with major problems. In this paper, we have proposed using the free Google Earth (GE) images which have high spatial resolution and high-coverage of land surface to enhance the spatial information of RapidEye images. A simulated panchromatic image has been generated by three band GE image and with three different methods: Mean, principal component analysis (PCA) and weighted average of GE image bands. In the last method, the weights are extracted from the spectral response curve of the satellite which captured the GE image. The simulated panchromatic image has been utilized for pansharpening of RapidEye image in five well-known methods: Brovey, Gram-Schmidt (GS), intensity-hue-saturation (IHS), Pansharp1 and Pansharp2. The most important point is finding the GE image with lowest lag time with RapidEye image. By satisfying this condition, the experiments illuminated that the proposed method can effectively enhance the spatial quality of RapidEye image. Also, this study presented that Pansharp2 method, which used simulated panchromatic image generated by the spectral response curve information, has revealed the best results of RapidEye image pansharpening.     

Monitoring deforestation and forest degradation using multi-temporal fraction images derived from Landsat sensor data in the Brazilian Amazon

Deforestation is the replacement of forest by other land use while degradation is a reduction of long-term canopy cover and/or forest stock. Forest degradation in the Brazilian Amazon is mainly due to selective logging of intact/un-managed forests and to uncontrolled fires. The deforestation contribution to carbon emission is already known but determining the contribution of forest degradation remains a challenge. Discrimination of logging from fires, both of which produce different levels of forest damage, is important for the UNFCCC (United Nations Framework Convention on Climate Change) REDD+ (Reducing Emissions from Deforestation and Forest Degradation) program. This work presents a semi-automated procedure for monitoring deforestation and forest degradation in the Brazilian Amazon using fraction images derived from Linear Spectral Mixing Model (LSMM). Part of a Landsat Thematic Mapper (TM) scene (path/row 226/068) covering part of Mato Grosso State in the Brazilian Amazon, was selected to develop the proposed method. First, the approach consisted of mapping deforested areas and mapping forest degraded by fires using image segmentation. Next, degraded areas due to selective logging activities were mapped using a pixel-based classifier. The results showed that the vegetation, soil, and shade fraction images allowed deforested areas to be mapped and monitored and to separate degraded forest areas caused by selective logging and by fires. The comparison of Landsat Operational Land Imager (OLI) and RapidEye results for the year 2013 showed an overall accuracy of 94%. We concluded that spatial resolution plays an important role for mapping selective logging features due to their characteristics. Therefore, when compared to Landsat data, the current availability of higher spatial and temporal resolution data, such as provided by Sentinel-2, is expected to improve the assessment of deforestation and forest degradation, especially caused by selective logging. This will facilitate the implementation of actions for forest protection.     

面向内陆水环境监测的GF-1卫星WFV数据特征评价

各种卫星遥感数据在内陆水环境监测中得到日益广泛的应用并各具优势和不足。2013年4月发射的高分一号(GF\|1)卫星搭载的宽视场相机(WFV)为水环境监测提供了新的数据源。通过与Landsat8 OLI和HJ\|1 CCD对比,从辐射、光谱和空间3个方面客观评价GF\|1 WFV 的数据特征,并分析其在内陆水环境监测应用中的优缺点。结果表明:WFV在内陆水体区域的灰度级数和信噪比高于CCD但低于OLI 10 bit的量化等级足以满足水质参数反演的精度要求,在水体监测应用中WFV现有辐射定标系数需修正;WFV的波段数量和宽度与CCD基本一致,比OLI的少且宽,不能很好地捕捉内陆水体的光谱特征;WFV16 m的空间分辨率和800 km的幅宽,明显优于CCD和OLI。总之,WFV在大范围中小型内陆水体环境动态监测方面具有良好的应用前景。     

融合不精准先验知识的Landsat 8 OLI影像深度学习分类方法

遥感影像解译是获得土地利用和土地覆盖（LULC）信息最为重要的途径之一,而自动化分类是提高LULC信息获取效率的关键。实际场景中包含大量不精准的先验知识,提取并融合其中的可用知识能进一步提高影像分类方法的精度、自动化率和规模应用能力。基于上述情况,提出了一种融合不精准先验知识的Landsat 8 OLI影像深度学习分类方法。该方法可自动规避先验知识中的不精准单元,在图斑约束空间内实现了分类样本的自动化区域选择和特征提取,并获得了高置信度知识,然后利用这些分类样本训练深度残差网络,从而实现大区域影像的精确分类。以常州市新北区为例进行实验,选用该区域2009年土地利用现状数据作为先验数据,2014年Landsat 8 OLI影像作为待分类影像。实验结果表明,所提方法可融合不精准先验知识,对大面积连片LULC信息分类精确,主要地类图斑界限准确,全图分类图斑精度达到了88.7%,Kappa系数为0.842。该方法能配合深度学习方法实现高精度Landsat 8 OLI遥感影像分类。     

融合不精准先验知识的Landsat 8 OLI影像深度学习分类方法

遥感影像解译是获得土地利用和土地覆盖（LULC）信息最为重要的途径之一，而自动化分类是提高LULC信息获取效率的关键。实际场景中包含大量不精准的先验知识，提取并融合其中的可用知识能进一步提高影像分类方法的精度、自动化率和规模应用能力。基于上述情况，提出了一种融合不精准先验知识的Landsat 8 OLI影像深度学习分类方法。该方法可自动规避先验知识中的不精准单元，在图斑约束空间内实现了分类样本的自动化区域选择和特征提取，并获得了高置信度知识，然后利用这些分类样本训练深度残差网络，从而实现大区域影像的精确分类。以常州市新北区为例进行实验，选用该区域2009年土地利用现状数据作为先验数据，2014年Landsat 8 OLI影像作为待分类影像。实验结果表明，所提方法可融合不精准先验知识，对大面积连片LULC信息分类精确，主要地类图斑界限准确，全图分类图斑精度达到了88.7%，Kappa系数为0.842。该方法能配合深度学习方法实现高精度Landsat 8 OLI遥感影像分类。     

基于Landsat 8卫星影像的北京地区土地覆盖分类

目的 土地覆盖分类能为生态系统模型、水资源模型和气候模型等提供重要信息,遥感技术运用于土地覆盖分类具有诸多优势。作为区域性土地覆盖分类应用的重要数据源,Landsat 5/7的TM和ETM+等数据已逐渐失效,Landsat 8陆地成像仪(OLI)较TM和ETM+增加了新的特性,利用Landsat 8数据进行北京地区土地覆盖分类研究,探讨处理方法的适用性。方法 首先,确定研究区域内土地覆盖分类系统,并对Landsat 8多光谱数据进行预处理,包括大气校正、地形校正、影像拼接及裁剪;然后,利用灰度共生矩阵提取全色波段纹理信息,与多光谱数据进行融合;最后,使用支持向量机(SVM)进行分类,获得土地覆盖分类结果。结果 经过精度评价和分析发现,6S模型大气校正和C模型地形校正预处理提高了不同类别之间的可分性,多光谱数据结合全色波段纹理特征能有效提高部分地物的土地覆盖分类精度,总体精度提高2.8%。结论 相对于Landsat TM/ETM+数据,Landsat 8 OLI数据新增特性有利于土地覆盖分类精度的提高。本文方法适用于Landsat 8 OLI数据土地覆盖分类研究与应用,能够满足大区域土地覆盖分类应用需求。     

Statistical model development and estimation of suspended particulate matter concentrations with Landsat 8 OLI images of Dongting Lake,China

Suspended particulate matter (SPM) is a dominant water constituent of case-II waters, and SPM concentration (C_SPM) is a key parameter describing water quality. This study, using Landsat 8 Operational Land Imager (OLI) images, aimed to develop the C_SPM retrieval models and further to estimate the C_SPM values of Dongting Lake. One Landsat 8 OLI image and 53 C_SPM measurements were employed to calibrate Landsat 8-based C_SPM retrieval models. The C_SPM values derived from coincident Landsat 8 OLI and Moderate Resolution Imaging Spectroradiometer (MODIS) images were compared to validate calibrated Landsat 8-based C_SPM models. After the best stable Landsat 8-based C_SPM retrieval model was further validated using an independent Landsat 8 OLI image and its coincident C_SPM measurements, it was applied to four Landsat 8 OLI images to retrieve the C_SPM values in the South and East Dongting Lake. Model calibration results showed that two exponential models of the red band explained 61% (estimated standard error (SE) = 7.96 mg l^–1) and 67% (SE = 6.79 mg l^–1) of the variation of C_SPM; two exponential models of the red:panchromatic band ratio obtained 81% (SE = 5.48 mg l^–1) and 77% (SE = 4.96 mg l^–1) fitting accuracy; and four exponential and quadratic models of the infrared band explained 72–83% of the variation of C_SPM (SE = 5.18–5.52 mg l^–1). By comparing the MODIS- and Landsat 8-based C_SPM values, an exponential model of the Landsat 8 OLI red band (C_SPM = 1.1034 × exp(23.61 × R)) obtained the best consistent C_SPM estimations with the MODIS-based model (r = 0.98, p < 0.01), and its further validation result using an independent Landsat 8 OLI image showed a significantly strong correlation between the measured and estimated C_SPM values at a significance level of 0.05 (r = 0.91, p < 0.05). The C_SPM spatiotemporal distribution derived from four Landsat 8 images revealed a clear spatial distribution pattern of C_SPM in the South and East Dongting Lake, which was caused by natural and anthropogenic factors together. This study confirmed the potential of Landsat 8 OLI images in retrieving C_SPM and provided a foundation for retrieving the spatial distribution of C_SPM accurately from this new data source in Dongting Lake.     

用高分一号数据提取玉米面积及精度分析

由于受到时间分辨率的影响，长期以来国内遥感技术在面积监测、作物长势监测等方面受到限制。针对此问题，该文利用“高分一号”卫星高空间和高时间分辨率的特点，应用其宽幅16m分辨率数据，结合Landsat 8和RapidEye数据，采用支持向量机（SVM）和光谱角法（SAM）在许昌进行农作物（玉米）的识别和面积提取及其精度分析。结果表明，“高分一号”4个宽幅传感器的影像应用精度差别较大，其中WFV3数据的作物识别与种植面积提取精度最高，高于Landsat 8，与RapidEye接近；而WFV1和WFV4数据的应用效果较差，不太适用于试验区内复杂的秋季作物类型的识别。总体上讲，SVM分类器的分类精度和Kappa系数都要好于SAM分类器，相比之下SVM更适合于农作物的识别和种植面积提取。     

基于多时相Landsat8 OLI影像的作物种植结构提取

针对基于多时相遥感影像、多种特征量提取多种作物种植结构在我国研究较少的现状,利用多时相Landsat8OLI影像数据,根据温宿县不同作物的农事历,通过分析主要地物的光谱特征和归一化植被指数的时间变化信息,构建不同作物种植结构提取的决策树模型,实现了对温宿县多种作物种植结构信息的提取。结果表明:1水稻的最佳识别依据是5月20日影像的近红外波段和7月23日影像的NDVI值;棉花和春玉米的最佳识别依据是5月20日~9月9日影像的NDVI变化值;冬小麦—夏玉米和林果的最佳识别依据是5月20日~7月23日影像的NDVI变化值;2与单时相监督分类相比,多时相决策树法对多种作物种植结构的提取效果更理想,总体精度提高了7.90%,Kappa系数提高了0.10;3Landsat8OLI影像数据分辨率高、成本低、获取方便,是农作物遥感的良好数据源。     

Satellite data cloud detection using deep learning supported by hyperspectral data

ABSTRACT

Deep learning methods can play an important role in satellite data cloud detection. The number and quality of training samples directly affect the accuracy of cloud detection based on deep learning. Therefore, selecting a large number of representative and high-quality training samples is a key step in cloud detection based on deep learning. For different satellite data sources, choosing sufficient and high-quality training samples has become an important factor limiting the application of deep learning in cloud detection. This paper presents a fast method for obtaining high-quality learning samples, which can be used for cloud detection of different satellite data with deep learning methods. AVIRIS (Airborne Visible Infrared Imaging Spectrometer) data, which have 224 continuous bands in the spectral range from 400–2500 nm, are used to provide cloud detection samples for different types of satellite data. Through visual interpretation, a sufficient number of cloud and clear sky pixels are selected from the AVIRIS data to construct a hyperspectral data sample library, which is used to simulate different satellite data (such as data from the Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat Operational Land Imager (OLI) satellites) as training samples. This approach avoids selecting training samples for different satellite sensors. Based on the Keras deep learning framework platform, a backpropagation (BP) neural network is employed for cloud detection from Landsat 8 OLI, National Polar-orbiting Partnership (NPP) Visible Infrared Imaging Radiometer Suite (VIIRS) and Terra MODIS data. The results are compared with cloud coverage results interpreted via artificial vision. The results demonstrate that the algorithm achieves good cloud detection results for the above data, and the overall accuracy is greater than 90%.     

Noise over water surfaces in Landsat TM images

A coherent pattern of system noise observable on all visible wavebands of Landsat Thematic Mapper images over homogeneous surfaces such as waterbodies is regarded as serious enough to impair visual interpretation and affect image analysis results. The noise is removable using iterative median filtering in the spatial domain, which requires much less processing time than removal by frequency domain Fourier transform. The significance of the error introduced to the images by the noise is evaluated in terms of water quality parameters in the study area, and methods for removal of the noise are described.