Identifying buildings from remote sensing imagery has been a challenge due to uncertainties from remote sensing imagery and variations in building structure and texture. In this study, we develop a scale robust CNN structure to improve the segmentation accuracy of building data from high-resolution aerial and satellite images. Based on a fully convolutional network, we introduce two Atrous convolutions on the first two lowest-scale layers, respectively, in the decoding step, aiming at enlarging the sight-of-view and integrate semantic information of large buildings. Then, a multi-scale aggregation strategy is applied. The last feature maps of each scale are used to predict the corresponding building labels, and further up-sampled to the original scale and concatenated for the final prediction. In addition, we introduce a combined data augmentation and relative radiometric calibration method for multi-source building extraction. The method enlarges sample spaces and hence the generalization ability of the deep learning models. We validate our developed methods with an aerial dataset of more than 180, 000 buildings with various architectural types, and a satellite image dataset consists of more than 29,000 buildings. The results are compared with several most recent studies. The comparison result shows our neural network outperformed other studies, especially in segmenting scenes of large buildings. The test on transfer learning from aerial dataset to satellite dataset showed our augmentation strategy significantly improved the prediction accuracy; however, further studies are needed to improve the generalization ability of the CNN model.
The study investigates land use/cover classification and change detection of urban areas from very high resolution (VHR) remote sensing images using deep learning-based methods. Firstly, we introduce a fully Atrous convolutional neural network (FACNN) to learn the land cover classification. In the FACNN an encoder, consisting of full Atrous convolution layers, is proposed for extracting scale robust features from VHR images. Then, a pixel-based change map is produced based on the classification map of current images and an outdated land cover geographical information system (GIS) map. Both polygon-based and object-based change detection accuracy is investigated, where a polygon is the unit of the GIS map and an object consists of those adjacent changed pixels on the pixel-based change map. The test data covers a rapidly developing city of Wuhan (8000 km2), China, consisting of 0.5 m ground resolution aerial images acquired in 2014, and 1 m ground resolution Beijing-2 satellite images in 2017, and their land cover GIS maps. Testing results showed that our FACNN greatly exceeded several recent convolutional neural networks in land cover classification. Second, the object-based change detection could achieve much better results than a pixel-based method, and provide accurate change maps to facilitate manual urban land cover updating.
Atmospheric mercury deposition by wet and dry processes contributes to the transformation of mercury from atmosphere to terrestrial and aquatic systems. Factors influencing the amount of mercury deposited to subtropical forests were identified in this study. Throughfall and open field precipitation samples were collected in 2012 and 2013 using precipitation collectors from forest sites located across Mt. Jinyun in southwest China. Samples were collected approximately every 2 weeks and analyzed for total (THg) and methyl mercury (MeHg). Forest canopy was the primary factor on THg and MeHg deposition. Simultaneously, continuous measurements of atmospheric gaseous elemental mercury (GEM) were carried out from March 2012 to February 2013 at the summit of Mt. Jinyun. Atmospheric GEM concentrations averaged 3.8 ± 1.5 ng m(-3), which was elevated compared with global background values. Sources identification indicated that both regional industrial emissions and long-range transport of Hg from central, northeast, and southwest China were corresponded to the elevated GEM levels. Precipitation deposition fluxes of THg and MeHg in Mt. Jinyun were slightly higher than those reported in Europe and North America, whereas total fluxes of MeHg and THg under forest canopy on Mt. Jiuyun were 3 and 2.9 times of the fluxes of THg in wet deposition in the open. Highly elevated litterfall deposition fluxes suggest that even in remote forest areas of China, deposition of atmospheric Hg(0) via uptake by vegetation leaf may be a major pathway for the deposition of atmospheric Hg. The result illustrates that areas with greater atmospheric pollution can be expected to have greater fluxes of Hg to soils via throughfall and litterfall.
The (p)ppGpp signal molecules play a central role in the stringent response (SR) to adapt to nutrient starvation in bacteria, yet the carbohydrate starvation induced adaptive response and the roles of SR in this response is not well characterized, especially in Gram-positives. Here, two (p)ppGpp synthetases RelA and RelQ are identified in Streptococcus suis, an important emerging zoonotic Gram-positive bacterium, while only RelA is functional under glucose starvation. To characterize the roles of RelA/(p)ppGpp in glucose starvation response in S. suis, the growth curves and transcriptional profiles were compared between the mutant strain ΔrelA [a (p)ppGpp0 strain under glucose starvation] and its parental strain SC-19 [(p)ppGpp+]. The results showed great difference between SC-19 and ΔrelA on adaptive responses when suffering glucose starvation, and demonstrated that RelA/(p)ppGpp plays important roles in adaptation to glucose starvation. Besides the classic SR including inhibition of growth and related macromolecular synthesis, the extended adaptive response also includes inhibited glycolysis, and carbon catabolite repression (CCR)-mediated carbohydrate-dependent metabolic switches. Collectively, the pheno- and genotypic characterization of the glucose starvation induced adaptive response in S. suis makes a great contribution to understanding better the mechanism of SR.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.