The task of instance segmentation in remote sensing images, aiming at performing per-pixel labeling of objects at the instance level, is of great importance for various civil applications. Despite previous successes, most existing instance segmentation methods designed for natural images encounter sharp performance degradations when they are directly applied to top-view remote sensing images. Through careful analysis, we observe that the challenges mainly come from the lack of discriminative object features due to severe scale variations, low contrasts, and clustered distributions. In order to address these problems, a novel context aggregation network (CATNet) is proposed to improve the feature extraction process. The proposed model exploits three lightweight plug-and-play modules, namely, dense feature pyramid network (DenseFPN), spatial context pyramid (SCP), and hierarchical region of interest extractor (HRoIE), to aggregate global visual context at feature, spatial, and instance domains, respectively. DenseFPN is a multi-scale feature propagation module that establishes more flexible information flows by adopting interlevel residual connections, cross-level dense connections, and feature reweighting strategy. Leveraging the attention mechanism, SCP further augments the features by aggregating global spatial context into local regions. For each instance, HRoIE adaptively generates RoI features for different downstream tasks. Extensive evaluations of the proposed scheme on iSAID, DIOR, NWPU VHR-10, and HRSID datasets demonstrate that the proposed approach outperforms state-of-the-arts under similar computational costs. Source code and pretrained models are available at https://github.com/yeliudev/CATNet.

In the task of change detection (CD), high-resolution remote sensing images (HRSIs) can provide rich ground object information. However, the interference from noise and complex background information can also bring some challenges to CD. In recent years, deep learning methods represented by convolutional neural networks (CNNs) have achieved good CD results. However, the existing methods have difficulty in detecting the detailed change information of the ground objects effectively. The imbalance of positive and negative samples can also seriously affect the CD results. In this letter, to solve the above problems, we propose a method based on a multiscale fully convolutional neural network (MFCN), which uses multiscale convolution kernels to extract the detailed features of the ground object features. A loss function combining weighted binary cross-entropy (WBCE) loss and dice coefficient loss is also proposed, so that the model can be trained from unbalanced samples. The proposed method was compared with six state-of-the-art CD methods on the DigitalGlobe dataset. The experiments showed that the proposed method can achieve a higher F1 -score, and the detection effect of the detailed changes was better than that of the other methods.

Visible remotely sensed images usually suffer from the haze, which contaminates the surface radiation and degrades the data quality in both spatial and spectral dimensions. This study proposes a spatial-spectral adaptive haze removal method for visible remote sensing images to resolve spatial and spectral problems. Spatial adaptation is considered from global and local aspects. A globally nonuniform atmospheric light model is constructed to depict spatially varied atmospheric light. Moreover, a bright pixel index is built to extract local bright surfaces for transmission correction. Spectral adaptation is performed by exploring the relationships between image gradients and transmissions among bands to estimate spectrally varied transmission. Visible remote sensing images featuring different land covers and haze distributions were collected for synthetic and real experiments. Accordingly, four haze removal methods were selected for comparison. Visually, the results of the proposed method are completely free from haze and colored naturally in all experiments. These outcomes are nearly the same as the ground truth in the synthetic experiments. Quantitatively, the mean-absolute-error, root-mean-square-error, and spectral angle are the smallest, and the coefficient-of-determination (R²) is the largest among the five methods in the synthetic experiments. R², structural similarity index measure, and the correlation coefficient between the result of the proposed method and the reference image are closest to 1 in the real data experiments. All experimental analyses demonstrate that the proposed method is effective in removing haze and recovering ground information faithfully under different scenes.

Downscaled microwave soil moisture (SM) products with a fine resolution are of great importance for both local and regional studies. However, few studies have explored the merits of multiple downscaled microwave SM products. An evaluation of the different products could help to advance knowledge of the downscaled microwave SM products and help researchers to choose the appropriate downscaled SM products for use in further studies. In this research, five microwave SM products derived from Advanced Microwave Scanning Radiometer–Earth Observing System (AMSR-E), AMSR2, and Soil Moisture and Ocean Salinity (SMOS) data were downscaled via the back-propagation neural network (BPNN). The BPNN was chosen because it can effectively simulate the nonlinear relationship between SM and the land surface temperature (LST)/vegetation index (VI). The different downscaled SM products were evaluated with in-situ SM data from the central Tibetan Plateau Soil Moisture/Temperature Monitoring Network (SMTMN) during the period from 1 August 2010 to 31 December 2012. Compared with the regression technique, the downscaled correlation coefficient (r) is significantly improved by the BPNN. The downscaled root-mean-square error (RMSE) and bias are comparable for the two techniques. As expected, LST and enhanced VI (EVI) are physically related to SM, and this is the most suitable combination for SM downscaling. Except for the ascending node of SMOS and AMSR2, the downscaled r is closely related to the original RMSE, and a lower original RMSE for the SM product results in a higher downscaled r. The BPNN-downscaled SMOS product in descending node is the closest to the in-situ SM among the different downscaled microwave SM products. The temporal variations and ranges of the microwave SM products are well maintained by the BPNN downscaling. Furthermore, the evaluations against in-situ SM reveal that the overall accuracies of the BPNN-downscaled SM products are very close to the original microwave SM products.