Image super-resolution (SR) techniques improve various remote sensing applications by allowing for finer spatial details than those captured by the original acquisition sensors. Recent advances in ...deep learning bring a new opportunity for SR by learning the mapping from low to high resolution. The most used convolutional neural networks (CNN) based approaches are prone to excessive smoothing or blurring due to the optimization objective in mean squared error (MSE). Instead, generative adversarial network (GAN) based approaches can achieve more perceptually acceptable results. However, the preliminary design of GANs generator with simple direct- or skip-connection residual blocks compromises its SR potential. Emerging dense convolutional network (DenseNet) equipped with dense connections has shown a promising prospect in classification and super-resolution. An intuitive idea to introduce DenseNet into GAN is expected to boost SR performance. However, because convolutional kernels in the existing residual block are arranged into a one-dimensional flat structure, the formation of dense connections highly relies on skip connections (linking the current layer to all subsequent layers with a shortcut path). In order to increase connection density, the depth of the layer has to be accordingly expanded, which in turn results in training difficulties such as vanishing gradient and information propagation loss. To this end, this paper proposes an ultra-dense GAN (udGAN) for image SR, where we reform the internal layout of the residual block into a two-dimensional matrix topology. This topology can provide additional diagonal connections so that we can still accomplish enough pathways with fewer layers. In particular, the pathways are almost doubled compared to previous dense connections under the same number of layers. The achievable rich connections are flexibly adapted to the diversity of image content, thus leading to improved SR performance. Extensive experiments on public benchmark datasets and real-world satellite imagery show that our model outperforms state-of-the-art counterparts in both subjective and quantitative assessments, especially those related to perception.
Multispectral satellite imaging sensors acquire various spectral band images and have a unique spectroscopic property in each band. Unfortunately, image artifacts from imaging sensor noise often ...affect the quality of scenes and have a negative impact on applications for satellite imagery. Recently, deep learning approaches have been extensively explored to remove noise in satellite imagery. Most deep learning denoising methods, however, follow a supervised learning scheme, which requires matched noisy image and clean image pairs that are difficult to collect in real situations. In this article, we propose a novel unsupervised multispectral denoising method for satellite imagery using a wavelet directional cycle-consistent adversarial network (WavCycleGAN). The proposed method is based on an unsupervised learning scheme using adversarial loss and cycle-consistency loss to overcome the lack of paired data. Moreover, in contrast to the standard image-domain cycleGAN, we introduce a wavelet directional learning scheme for effective denoising without sacrificing high-frequency components such as edges and detailed information. Experimental results for the removal of vertical stripes and wave noise in satellite imaging sensors demonstrate that the proposed method effectively removes noise and preserves important high-frequency features of satellite images.
Nigerian health officials won't have to rely on flawed, decade-old census data when they plan deliveries of the measles vaccine next year. Instead, they will have access to what may be the most ...detailed and up-to-date population map ever produced for a developing country. Created by the Bill & Melinda Gates Foundation in Seattle, Washington, and delivered to Nigerian officials on 1 May, the map is based on a detailed analysis of buildings in satellite imagery and more than 2,000 on-the-ground neighbourhood surveys.
•An analytical framework is developed to detect probable illegal dump sites (IDSs)•An IDS probability map is produced at a 31,285 km2 site in rural and remote regions.•Populated areas represented by ...nighttime light is a priori of IDS.•Among all classes, highway length is the most decisive factor on IDS probability.•Vulnerability of the inhabitants on reserve lands was assessed in 3 sample regions.
Illegal dump sites (IDS) pose significant risks to human and the environment and are a pressing issue worldwide. Due to their secretive nature, the detection of IDS is costly and ineffective. In this study, an analytical framework was developed to detect probable IDSs in rural and remote areas using nighttime light (NTL) as a proxy for populated areas. An IDS probability map is produced by aggregation of Landsat-8 and Suomi NPP satellite imagery, multiple-criteria decision-making analysis, and classification tools. Six variables are considered, including modified soil adjusted index, land surface temperature, NTL, highway length, railway length, and the number of landfills. Vulnerability of the inhabitants on reserve lands was assessed using three sample regions. The method appears effective in reducing potential IDSs. Only about 7% of the 31,285 km2 study area are identified as probable IDS, being classified as “very high” and “high”. Landfills without permit are found more effective in lowering IDS occurrence. Spatial distributions of reserve lands and the maturity of highways network nearby may be more important than the length of railways when assessing the inhabitant vulnerability due to IDS. Highway length is the most decisive factor on IDS probability among all classes, with membership grades ranging from 0.99 to 0.55. Land surface temperature appears less effective for the identification of smaller scale IDS. NTL is more prominent on IDS probability in the “very high” class, with a membership grade of 0.80. The finding suggests that populated areas represented by NTL is a priori of IDS.
The two Moderate Resolution Imaging Spectroradiometer (MODIS) instruments on-board NASA's Terra and Aqua satellites have provided nearly two decades of global fire data. Here, we describe refinements ...made to the 500-m global burned area mapping algorithm that were implemented in late 2016 as part of the MODIS Collection 6 (C6) land-product reprocessing. The updated algorithm improves upon the heritage Collection 5.1 (C5.1) MCD64A1 and MCD45A1 algorithms by offering significantly better detection of small burns, a modest reduction in burn-date temporal uncertainty, and a large reduction in the extent of unmapped areas. Comparison of the C6 and C5.1 MCD64A1 products for fifteen years (2002–2016) on a regional basis shows that the C6 product detects considerably more burned area globally (26%) and in almost every region considered. The sole exception was in Boreal North America, where the mean annual area burned was 6% lower for C6, primarily as a result of a large increase in the number of small lakes mapped (and subsequently masked) at high latitudes in the upstream C6 input data. With respect to temporal reporting accuracy, 44% of the C6 MCD64A1 burned grid cells were detected on the same day as an active fire, and 68% within 2 days, which represents a substantial reduction in temporal uncertainty compared to the C5.1 MCD64A1 and MCD45A1 products. In addition, an areal accuracy assessment of the C6 burned area product undertaken using high resolution burned area reference maps derived from 108 Landsat image pairs is reported.
•We describe the MODIS Collection 6 (C6) MCD64A1 burned area mapping algorithm.•The C6 algorithm improves upon the heritage Collection 5.1 (C5.1) MODIS algorithms.•The C6 algorithm offers significantly better detection of small burns.•The C6 product detects 26% more burned area globally.•We performed a Stage-2 validation of the C5.1 and C6 MCD64A1 products.
•Summarize deep learning methods for semantic segmentation of remote sensing images.•Identify three major challenges faced by researchers.•Summarize the innovative development to address them.
...Semantic segmentation of remote sensing imagery has been employed in many applications and is a key research topic for decades. With the success of deep learning methods in the field of computer vision, researchers have made a great effort to transfer their superior performance to the field of remote sensing image analysis. This paper starts with a summary of the fundamental deep neural network architectures and reviews the most recent developments of deep learning methods for semantic segmentation of remote sensing imagery including non-conventional data such as hyperspectral images and point clouds. In our review of the literature, we identified three major challenges faced by researchers and summarize the innovative development to address them. As tremendous efforts have been devoted to advancing pixel-level accuracy, the emerged deep learning methods demonstrated much-improved performance on several public data sets. As to handling the non-conventional, unstructured point cloud and rich spectral imagery, the performance of the state-of-the-art methods is, on average, inferior to that of the satellite imagery. Such a performance gap also exists in learning from small data sets. In particular, the limited non-conventional remote sensing data sets with labels is an obstacle to developing and evaluating new deep learning methods.
Large‐bodied animals such as baleen whales can now be detected with very high resolution (VHR) satellite imagery, allowing for scientific studies of whales in remote and inaccessible areas where ...traditional survey methods are limited or impractical. Here we present the first study of baleen whales using the WorldView‐3 satellite, which has a maximum spatial resolution of 31 cm in the panchromatic band, the highest currently available to nonmilitary professionals. We manually detected, described, and counted four different mysticete species: fin whales (Balaenoptera physalus) in the Ligurian Sea, humpback whales (Megaptera novaeangliae) off Hawaii, southern right whales (Eubalaena australis) off Península Valdés, and gray whales (Eschrichtius robustus) in Laguna San Ignacio. Visual and spectral analyses were conducted for each species, their surrounding waters, and nonwhale objects (e.g., boats). We found that behavioral and morphological differences made some species more distinguishable than others. Fin and gray whales were the easiest to discern due to their contrasting body coloration with surrounding water, and their prone body position, which is proximal to the sea surface (i.e., body parallel to the sea surface). These results demonstrate the feasibility of using VHR satellite technology for monitoring the great whales.
The interactions between climate and land‐use change are dictating the distribution of flora and fauna and reshuffling biotic community composition around the world. Tropical mountains are ...particularly sensitive because they often have a high human population density, a long history of agriculture, range‐restricted species, and high‐beta diversity due to a steep elevation gradient. Here we evaluated the change in distribution of woody vegetation in the tropical Andes of South America for the period 2001–2014. For the analyses we created annual land‐cover/land‐use maps using MODIS satellite data at 250 m pixel resolution, calculated the cover of woody vegetation (trees and shrubs) in 9,274 hexagons of 115.47 km2, and then determined if there was a statistically significant (p < 0.05) 14 year linear trend (positive—forest gain, negative—forest loss) within each hexagon. Of the 1,308 hexagons with significant trends, 36.6% (n = 479) lost forests and 63.4% (n = 829) gained forests. We estimated an overall net gain of ~500,000 ha in woody vegetation. Forest loss dominated the 1,000–1,499 m elevation zone and forest gain dominated above 1,500 m. The most important transitions were forest loss at lower elevations for pastures and croplands, forest gain in abandoned pastures and cropland in mid‐elevation areas, and shrub encroachment into highland grasslands. Expert validation confirmed the observed trends, but some areas of apparent forest gain were associated with new shade coffee, pine, or eucalypt plantations. In addition, after controlling for elevation and country, forest gain was associated with a decline in the rural population. Although we document an overall gain in forest cover, the recent reversal of forest gains in Colombia demonstrates that these coupled natural‐human systems are highly dynamic and there is an urgent need of a regional real‐time land‐use, biodiversity, and ecosystem services monitoring network.
The interactions between climate and land‐use change are dictating the distribution of flora and fauna and reshuffling biotic community composition around the world. In the Andes of South America, we found a net increase in woody vegetation above 1000 m. While climate change has likely contributed to this increase, especially at higher elevations, land‐use change is the primary factor altering the contemporary distributions of many species.
We report on the first global census of the Adélie Penguin (Pygoscelis adeliae), achieved using a combination of ground counts and satellite imagery, and find a breeding population 53% larger (3.79 ...million breeding pairs) than the last estimate in 1993. We provide the first abundance estimates for 41 previously unsurveyed colonies, which collectively contain 420,000 breeding pairs, and report on 17 previously unknown colonies, 11 of which may be recent colonizations. These recent colonizations represent ∼5% of the increase in known breeding population and provide insight into the ability of these highly philopatric seabirds to colonize new breeding territories. Additionally, we report on 13 colonies not found in the survey, including 8 that we conclude have gone extinct. We find that Adélie Penguin declines on the Antarctic Peninsula are more than offset by increases in East Antarctica. Our global population assessment provides a robust baseline for understanding future changes in abundance and distribution. These results are a critically needed contribution to ongoing negotiations regarding the design and implementation of Marine Protected Areas for the Southern Ocean.
Changes in vegetation cover associated with the observed greening may affect several biophysical processes, whose net effects on climate are unclear. We analyzed remotely sensed dynamics in leaf area ...index (LAI) and energy fluxes in order to explore the associated variation in local climate. We show that the increasing trend in LAI contributed to the warming of boreal zones through a reduction of surface albedo and to an evaporation-driven cooling in arid regions. The interplay between LAI and surface biophysics is amplified up to five times under extreme warm-dry and cold-wet years. Altogether, these signals reveal that the recent dynamics in global vegetation have had relevant biophysical impacts on the local climates and should be considered in the design of local mitigation and adaptation plans.