Search

Considering the classification of high spatial resolution remote sensing imagery, this paper presents a novel classification method for such imagery using deep neural networks. Deep learning methods, such as a fully convolutional network (FCN) model, achieve state-of-the-art performance in natural image semantic segmentation when provided with large-scale datasets and respective labels. To use data efficiently in the training stage, we first pre-segment training images and their labels into small patches as supplements of training data using graph-based segmentation and the selective search method. Subsequently, FCN with atrous convolution is used to perform pixel-wise classification. In the testing stage, post-processing with fully connected conditional random fields (CRFs) is used to refine results. Extensive experiments based on the Vaihingen dataset demonstrate that our method performs better than the reference state-of-the-art networks when applied to high-resolution remote sensing imagery classification. [ABSTRACT FROM AUTHOR]

Despite continuous improvements in microwave sensors and retrieval algorithms, our understanding of precipitation uncertainty is quite limited, due primarily to inconsistent findings in studies that compare satellite estimates toin situobservations over different parts of the world. This study seeks to characterize the temporal and spatial properties of uncertainty in the Tropical Rainfall Measuring Mission Microwave Imager surface rainfall product over tropical ocean basins. Two uncertainty analysis frameworks are introduced to qualitatively evaluate the properties of uncertainty under a hierarchy of spatiotemporal data resolutions. The first framework (i.e. ‘climate method’) demonstrates that, apart from random errors and regionally dependent biases, a large component of the overall precipitation uncertainty is manifested in cyclical patterns that are closely related to large-scale atmospheric modes of variability. By estimating the magnitudes of major uncertainty sources independently, the climate method is able to explain 45–88% of the monthly uncertainty variability. The percentage is largely resolution dependent (with the lowest percentage explained associated with a 1° × 1° spatial/1 month temporal resolution, and highest associated with a 3° × 3° spatial/3 month temporal resolution). The second framework (i.e. ‘weather method’) explains regional mean precipitation uncertainty as a summation of uncertainties associated with individual precipitation systems. By further assuming that self-similar recurring precipitation systems yield qualitatively comparable precipitation uncertainties, the weather method can consistently resolve about 50% of the daily uncertainty variability, with only limited dependence on the regions of interest. [ABSTRACT FROM PUBLISHER]

Due to advances in satellite and sensor technology, the number and size of Remote Sensing (RS) images continue to grow at a rapid pace. The continuous stream of sensor data from satellites poses major challenges for the retrieval of relevant information from those satellite datastreams. The Bag-of-Words (BoW) framework is a leading image search approach and has been successfully applied in a broad range of computer vision problems and hence has received much attention from the RS community. However, the recognition performance of a typical BoW framework becomes very poor when the framework is applied to application scenarios where the appearance and texture of images are very similar. In this paper, we propose a simple method to improve recognition performance of a typical BoW framework by representing images with local features extracted from base images. In addition, we propose a similarity measure for RS images by counting the number of same words assigned to images. We compare the performance of these methods with a typical BoW framework. Our experiments show that the proposed method has better recognition performance than that of the BoW and requires less storage space for saving local invariant features. [ABSTRACT FROM PUBLISHER]

This paper presents various limitations of the current remote sensing data distribution models and proposes a new concept called the location-based instant satellite image service for a new generation of remote sensing image distribution system. The essential feature of the service is that customers can subscribe to data based on the location of interest and satellite image data received by antenna will be distributed to customer's terminal devices instantly after imaging over the subscribed area. The workflow, architecture, and key technologies of the new generation data distribution system are described. The system is composed of four parts: data comprehensive processing component, data management component, product distribution component, and data display component. Based on this, a prototype system is developed, which demonstrates the promising service model with great potential for increased usage in many applications. [ABSTRACT FROM AUTHOR]

The cement industry, as one of the primary contributors to global greenhouse gas emissions, accounts for 7% of the world's carbon dioxide emissions. There is an urgent need to establish a rapid method for detecting cement plants to facilitate effective monitoring. In this study, a comprehensive method based on YOLOv5-IEG and the Thermal Signature Detection module using Google Earth optical imagery and SDGSAT-1 thermal infrared imagery was proposed to detect large-scale cement plant information, including geographic location and operational status. The improved algorithm demonstrated an increase of 4.8% in accuracy and a 7.7% improvement in MAP@.5:95. In a specific empirical investigation in China, we successfully detected 781 large-scale cement plants with an accuracy of 90.8%. Specifically, of the 55 cement plants in Shandong Province, we identified 46 as operational and nine as non-operational. The successful application of advanced models and remote sensing technology in efficiently and accurately tracking the operational status of cement plants provides crucial support for environmental protection and sustainable development. [ABSTRACT FROM AUTHOR]

Due to the limitation of technology and budget, it is often difficult for sensors of a single remote sensing satellite to have both high-temporal and high-spatial (HTHS) resolution at the same time. In this article, we proposed a new multilevel feature fusion with generative adversarial network (MLFF-GAN) for generating fusion HTHS images. The MLFF-GAN mainly uses U-net-like architecture, and its generator is composed of three stages: feature extraction, feature fusion, and image reconstruction. In the feature extraction and reconstruction stage, the generator employs the encoding and decoding structure to extract three groups of multilevel features (MLFs), which can cope with the huge difference of resolution between high-resolution images and low-resolution images. In the feature fusion stage, adaptive instance normalization (AdaIN) block is designed to learn the global distribution relationship between multitemporal images, and an attention module (AM) is used to learn the local information weights for the change of small areas. The proposed MLFF-GAN was tested on two Landsat and Moderate-Resolution Imaging Spectroradiometer (MODIS) datasets. Some state-of-the-art algorithms are comprehensively compared with MLFF-GAN. We also carried on the ablation experiment to test the effectiveness of different submodules in the MLFF-GAN. The experiment results and ablation analysis show the better performances of the proposed method when compared with other methods. The code is available at https://github.com/songbingze/MLFF-GAN. [ABSTRACT FROM AUTHOR]

With the rapid development of satellite remote sensing technology and an ever-increasing number of Earth observation satellites being launched, the global volume of remotely sensed imagery has been growing exponentially. Processing the variety of remotely sensed data has increasingly been complex and difficult. It is also hard to efficiently and intelligently retrieve what users need from a massive database of images. This paper introduces an improved support vector machine (SVM) model, which optimizes the model parameters and selects the feature subset based on the particle swarm optimization (PSO) method and genetic algorithm (GA) for remote sensing image retrieval. The results from an image retrieval experiment show that our method outperforms traditional methods such as GRID, PSO, and GA in terms of consistency and stability. [ABSTRACT FROM PUBLISHER]

Burn scar extraction using remote sensing data is an efficient way to precisely evaluate burn area and measure vegetation recovery. Traditional burn scar extraction methodologies have no well effect on burn scar image with blurred and irregular edges. To address these issues, this paper proposes an automatic method to extract burn scar based on Level Set Method (LSM). This method utilizes the advantages of the different features in remote sensing images, as well as considers the practical needs of extracting the burn scar rapidly and automatically. This approach integrates Change Vector Analysis (CVA), Normalized Difference Vegetation Index (NDVI) and the Normalized Burn Ratio (NBR) to obtain difference image and modifies conventional Level Set Method Chan-Vese (C-V) model with a new initial curve which results from a binary image applying K-means method on fitting errors of two near-infrared band images. Landsat 5 TM and Landsat 8 OLI data sets are used to validate the proposed method. Comparison with conventional C-V model, OSTU algorithm, Fuzzy C-mean (FCM) algorithm are made to show that the proposed approach can extract the outline curve of fire burn scar effectively and exactly. The method has higher extraction accuracy and less algorithm complexity than that of the conventional C-V model. [ABSTRACT FROM AUTHOR]

Considering the complexity of landslide hazards, their manual investigation lacks efficiency and is time-consuming, especially in high-altitude plateau areas. Therefore, extracting landslide information using remote sensing technology has great advantages. In this study, comprehensive research was carried out on the landslide features of high-resolution remote sensing images on the Mangkam dataset. Based on the idea of feature-driven classification, the landslide extraction model of a fully convolutional spectral–topographic fusion network (FSTF-Net) based on a deep convolutional neural network of multi-source data fusion is proposed, which takes into account the topographic factor (slope and aspect) and the normalized difference vegetation index (NDVI) as multi-source data input by which to train the model. In this paper, a high-resolution remote sensing image classification method based on a fully convolutional network was used to extract the landslide information, thereby realizing the accurate extraction of the landslide and surrounding ground-object information. With Mangkam County in the southeast of the Qinghai–Tibet Plateau China as the study area, the proposed method was evaluated based on the high-precision digital elevation model (DEM) generated from stereoscopic images of Resources Satellite-3 and multi-source high-resolution remote sensing image data (Beijing-2, Worldview-3, and SuperView-1). Results show that our method had a landslide detection precision of 0.85 and an overall classification accuracy of 0.89. Compared with the latest DeepLab_v3+, our model increases the landslide detection precision by 5%. Thus, the proposed FSTF-Net model has high reliability and robustness. [ABSTRACT FROM AUTHOR]

Class imbalance is a key issue for the application of deep learning for remote sensing image classification because a model generated by imbalanced samples training has low classification accuracy for minority classes. In this study, an accurate classification approach using the multistage sampling method and deep neural networks was proposed to classify imbalanced data. We first balance samples by multistage sampling to obtain the training sets. Then, a state-of-the-art model is adopted by combining the advantages of atrous spatial pyramid pooling (ASPP) and Encoder-Decoder for pixel-wise classification, which are two different types of fully convolutional networks (FCNs) that can obtain contextual information of multiple levels in the Encoder stage. The details and spatial dimensions of targets are restored using such information during the Decoder stage. We employ four deep learning-based classification algorithms (basic FCN, FCN-8S, ASPP, and Encoder-Decoder with ASPP of our approach) on multistage training sets (original, MUS1, and MUS2) of WorldView-3 images in southeastern Qinghai-Tibet Plateau and GF-2 images in northeastern Beijing for comparison. The experiments show that, compared with existing sets (original, MUS1, and identical) and existing method (cost weighting), the MUS2 training set of multistage sampling significantly enhance the classification performance for minority classes. Our approach shows distinct advantages for imbalanced data. [ABSTRACT FROM AUTHOR]

Ice storms greatly affect the structure, dynamics, and functioning of forest ecosystems. Studies on the impact of such disasters, as well as the post-disaster recovery of forests, are important contents in forest biology, ecology, and geography. Remote-sensing technology provides data and methods that can support the study of disasters at the large-to-medium scale and over long time periods. This study took Chebaling National Nature Reserve in Guangdong Province, China, as the study area. First, field-survey data and remote-sensing data were comprehensively analyzed to demonstrate the feasibility of replacing the forest stock volume with the mean annual value of the Enhanced Vegetation Index (EVI), to study forest growth and change. We then used the EVI from 2007 to 2017, together with a variety of other remote-sensing and forest sub-compartment data, to analyze the impact of the 2008 ice storm and the subsequent post-disaster recovery of the forest. Finally, we drew the following conclusions: (1) Topography had a considerable effect on disaster impact and forest recovery in Chebaling. The forest at high altitudes (700–1000 m) and on steep slopes (25–40°) was seriously affected by this disaster but had a stronger post-disaster recovery ability. Meanwhile, the hardest-hit area for coniferous forest was higher and steeper than that for broad-leaved forest. (2) In the same terrain conditions, coniferous forests were less affected by the disaster than broad-leaved forests and showed less variation during the post-disaster recovery process. Nevertheless, broad-leaved forests had faster recovery rates and higher recovery degrees; (3) Under the influence of human activities, the recovery and fluctuation degree for planted forest in the post-disaster recovery process was significantly higher than that for natural forest. The study suggests that forest has high disaster resistance and self-recovery ability after the ice storm, and this ability has a strong correlation with the type of forest and the topographic factors such as elevation and slope. At the same time, human intervention can speed up the recovery of forests after disasters. [ABSTRACT FROM AUTHOR]