Search

Building extraction from high-resolution remote sensing images is of great importance for urban planning, disaster assessment, and geography mapping. In recent years, convolutional neural networks have made outstanding achievements in improving the precision of building extraction. However, most existing approaches have some problems, such as insufficient detailed feature extraction and ignorance of the relationship between different features. In this study, we propose a novel multi-channel recurrent attention network (MCANet) for building extraction. Firstly, the multi-scale channel attention mechanism is used to expand the convolution kernel receptive field, making the model can extract rich building region feature information. Secondly, we use the spatial pyramid recurrent block to establish long-range dependencies over space, channel, and layer of different convolutions. Finally, the multi-channel feature fusion block is used to fuse the multi-scale channel features information, and improve the building extraction precision. Experimental results show that the proposed MCANet achieves better results (recall, precision, intersection-over-union, and F1_score on the Inria Aerial Image Labeling Dataset are 89.82%, 94.38%, 87.42%, and 88.25%, respectively), and outperforms the other state-of-the-art approaches.

Semantic segmentation of high-resolution remote sensing images has a wide range of applications, such as territorial planning, geographic monitoring and smart cities. The proper operation of semantic segmentation for remote sensing images remains challenging due to the complex and diverse transitions between different ground areas. Although several convolution neural networks (CNNs) have been developed for remote sensing semantic segmentation, the performance of CNNs is far from the expected target. This study presents a deep feature aggregation network (DFANet) for remote sensing image semantic segmentation. It is composed of a basic feature representation layer, an intermediate feature aggregation layer, a deep feature aggregation layer and a feature aggregation module (FAM). Specially, the basic feature representation layer is used to obtain feature maps with different resolutions: the intermediate feature aggregation layer and deep feature aggregation layer can fuse various resolution features and multi-scale features; the FAM is used to splice the features and form more abundant spatial feature maps; and the conditional random field module is used to optimize semantic segmentation results. We have performed extensive experiments on the ISPRS two-dimensional Vaihingen and Potsdam remote sensing image datasets and compared the proposed method with several variations of semantic segmentation networks. The experimental results show that DFANet outperforms the other state-of-the-art approaches.

An ultra-wideband and high-efficiency reflective linear-to-circular polarization conversion metasurface is proposed. The proposed metasurface is composed of a square array of a corner-truncated square patch printed on grounded dielectric substrate and covered with a dielectric layer, which is an orthotropic anisotropic structure with a pair of mutually perpendicular symmetric axes u and v along the directions with the tilt angles of ±45° with respect to the vertical y axis. When the u- and v-polarized waves are incident on the proposed metasurface, the phase difference between the two reflection coefficients is close to –90° in an ultra-wide frequency band, so it can realize high-efficiency and ultra-wideband LTC polarization conversion under both x- and y-polarized incidences in this band. The proposed polarization conversion metasurface is simulated and measured. Both the simulated and measured results show that the axial ratio (AR) of the reflected wave is kept below 3 dB in the ultra-wide frequency band of 5.87 GHz–21.13 GHz, which is corresponding to a relative bandwidth of 113%; moreover, the polarization conversion rate (PCR) can be kept larger than 99% in a frequency range of 8.08 GHz–20.92 GHz.

Location information is an important component of big data, so it has attracted many researchers’ attention to obtain the location of the target based on edge computing. However, in the actual application scenario, due to the influence of non-line-of-sight (NLOS) propagation, relative localization of anchorless networks in complex scenarios is a challenging issue. In this paper, we propose a robust localization framework based on a distributed architecture, which is suitable for the edge computing environment. The proposed framework consists of three steps, firstly, using hierarchical clustering method, the network is divided into several sub-clusters with a small number of nodes. Secondly, in each sub-cluster, outlier detection and low-rank matrix completion algorithms are used to complete the Euclidian distance matrix (EDM), furthermore, MDS is used to calculate the relative coordinates. In the last step, the relative coordinates of all sub-clusters are transformed and stitched to realize the whole network localization. In order to verify the effectiveness of the proposed framework, we carry out a large number of numerical simulations, the results show that our framework can effectively eliminate the outliers caused by NLOS and improve the localization performance in a complex environment.

Using the first principle method, we investigate the influence of Sr/Ti ratio on the atomic structure and dislocation behavior of Srn+1TinO3n+1(Ruddlesden—Popper phase) and SrnTin+1O3n+2 (Magneli phase). A linear lattice expansion versus the Sr/Ti ratio exhibits in the Ruddlesden—Popper and Magneli phases. The Ruddlesden—Popper phase has lower formation energy and superior structural stability than the Magneli phase. The two phases show different dislocation behaviors and it is found that a possibly preferred slip system 〈110〉 {110} emerges in the two phases, and the dislocations are more likely to dissociate into partial dislocations in Magneli phases.

Lowering the L10 ordering temperature of FePt films is of great significance for their application as an ultrahigh density magnetic recording medium. In this study, the L10 ordering process of FePt thin films deposited directly on Si substrates has been significantly accelerated by the interface reaction between the thin film and the Si substrate, and thus the thin films show a low L10 ordering temperature of T = 310??C as compared with those deposited on Si/SiO2 substrates. The accelerated L10 ordering transition is predominantly dependent on the rapid growth of the ordered domains during the interface reaction. The film thickness has an important effect on the interface reaction and thus can be used to tune the L10 ordering process of the FePt films.