Search

The information of cropland is obtained efficiently and accurately as the basis for achieving precision agriculture (PA). As the boundary between cropland and other types of land in remote sensing images with different resolutions is fuzzy, the characteristics of cropland are easily confused when extracting cropland, resulting in inaccurate identification and extraction of cropland under large and complex backgrounds and rough localization of marginal areas. We proposed a two-path multiscale attention self-supervised network with the perception of four directions in pixel space, called the multiscale bilateral spatial direction-aware network (MBSDANet), to solve these problems and improve the model’s ability to extract cropland in small samples. One path extracts attentional feature maps by spatial directions to preserve detailed direction-aware information and generate high-resolution features; the other path obtains local-to-global information through pyramid pooling and attention awareness to capture dense multiscale cropland features to separate targets in complex contexts. The features of the two branches at different levels are fused by weighting the multi-aware information. Triangular self-supervised and boundary-aware losses are used to achieve fine segmentation and extraction of cropland in small samples. We tested the extraction method on cropland in Denmark and the Hebei Province of China, demonstrating its effectiveness and generalization. Compared to other neural network models, MBSDANet achieves better accuracy with a precision of 0.9481, an IoU of 0.8937, and an F1 score of 0.9438.

The detection of small objects in Unmanned Aerial Vehicle (UAV) images, which contain a large number of objects with extremely small pixels, has been a major challenge over the last few decades. Owing to the lack of sufficiently detailed representation of object features, conventional detectors based on deep learning are unsatisfactory when detecting small objects. Meanwhile, the flight altitude and the shooting angle of the UAV always changing, which leads to an uneven dispersion of the multiple objects and a varying density of small objects. In response to the issues mentioned above, a novel lightweight Cross-layer Triple-branch Parallel Fusion Network (CTPFNet) is proposed to improve the real-time detection accuracy of small objects under UAV images. Firstly, a novel downsampling structure called the Inverse Residual Pooling Cascade (IRPC) module is proposed to obtain richer feature information about small objects. Secondly, we construct an improved global feature extraction structure Efficient Layer Aggregation Networks-Transformer (ELAN-Trans) to enhance the association between global features. Then, we design the Hybrid Dilated Depth Separable-Spatial Pyramid Pooling-Fast (HDDS-SPPF) to capture more contextual information by using the dilated convolution operations. To enhance the cross-scale transfer fusion of low-level original features containing more detailed localization information and deep-level deep features containing more semantic information, we propose a CTPF module embedded into the neck region for the secondary feature reuse of adjacent feature maps with varying resolutions. Extensive experiments on the public VisDrone2021-DET dataset show that the proposed model achieves significant performance gains with fewer parameters.

ABSTRACTDriven by improvements in satellite internet and Low Earth Orbit (LEO) navigation augmentation, the integration of communication and navigation has become increasingly common, and further improving navigation capabilities based on communication constellations has become a significant challenge. In the context of the existing Orthogonal Frequency Division Multiplexing (OFDM) communication systems, this paper proposes a new ranging signal design method based on an LEO satellite communication constellation. The LEO Satellite Communication Constellation Block-type Pilot (LSCC-BPR) signal is superimposed on the communication signal in a block-type form and occupies some of the subcarriers of the OFDM signal for transmission, thus ensuring the continuity of the ranging pilot signal in the time and frequency domains. Joint estimation in the time and frequency domains is performed to obtain the relevant distance value, and the ranging accuracy and communication resource utilization rate are determined. To characterize the ranging performance, the Root Mean Square Error (RMSE) is selected as an evaluation criterion. Simulations show that when the number of pilots is 2048 and the Signal-to-Noise Ratio (SNR) is 0 dB, the ranging accuracy can reach 0.8 m, and the pilot occupies only 50% of the communication subcarriers, thus improving the utilization of communication resources and meeting the public demand for communication and location services.

Abstract Most deep‐learning‐based multi‐view stereo series studies are concerned with improving the depth prediction accuracy of noise‐free images. However, it is difficult to obtain off‐the‐set clean images in practice and 3D convolutional neural networks require a lot of computing resources. To make full use of its computing power, different types of information can be processed simultaneously in the network. For these two issues, this paper proposes a novel multi‐stage network architecture to address depth inference and denoising simultaneously. Specifically, 2D feature maps are first converted into 3D cost volumes containing pixel information and depth information through differentiable homography and Gaussian probability mapping. Then, the cost volume is input into the regularisation module in each network stage to obtain the predicted probability volumes. Furthermore, simple static weights lead to training failure, and it is necessary to dynamically adjust the loss function by gradient normalisation. The proposed method can dispose of pixel information and depth information simultaneously and both reach an excellent level. Extensive experimental results show that the authors’ work surpasses the state‐of‐the‐art denoising on the DTU dataset (adding Gaussian–Poisson noise) and is more robust to noise images in depth inference.

In ocean remote sensing missions, recognizing an underwater acoustic target is a crucial technology for conducting marine biological surveys, ocean explorations, and other scientific activities that take place in water. The complex acoustic propagation characteristics present significant challenges for the recognition of underwater acoustic targets (UATR). Methods such as extracting the DEMON spectrum of a signal and inputting it into an artificial neural network for recognition, and fusing the multidimensional features of a signal for recognition, have been proposed. However, there is still room for improvement in terms of noise immunity, improved computational performance, and reduced reliance on specialized knowledge. In this article, we propose the Residual Attentional Convolutional Neural Network (RACNN), a convolutional neural network that quickly and accurately recognize the type of ship-radiated noise. This network is capable of extracting internal features of Mel Frequency Cepstral Coefficients (MFCC) of the underwater ship-radiated noise. Experimental results demonstrate that the proposed model achieves an overall accuracy of 99.34% on the ShipsEar dataset, surpassing conventional recognition methods and other deep learning models.

In agricultural remote sensing monitoring, climate often affects the quality of optical remote sensing image data acquisition. The acquired satellite imagery results usually contain cloud information, leading to a lack of ground data information. Unlike thick clouds, the semi-transparent nature of thin clouds prevents thin clouds from completely obscuring the ground scene. In order to remove thin clouds in the cultivated land and restore the actual ground information as much as possible, we proposed a cloud removal method of spatial information fusion self-attention generative adversarial network (SI-SA GAN) based on multi-directional perceptual attention and self-attention mechanism. The proposed method identifies and focuses on cloud regions using spatial attention, channel attention, and self-attention mechanism, which can enhance image information. The modules of the discriminator utilize residual networks and self-attention non-local neural networks to guide image information output. The generative adversarial network (GAN) is applied to remove clouds and restore the corresponding irregular occlusion area according to the depth characteristics of the input information. A gradient penalty is applied to improve the robustness of the generative network. In this paper, we compared the evaluation indexes of other advanced models. The qualitative and quantitative results of Sentinel-2A and public RICE datasets confirmed that the proposed method could enhance image quality effectively after cloud removal. The model has excellent thin cloud removal performance with small-scale training data.

Since the fingerprints of numerous crucial biologic materials can be identified by terahertz (THz) spectroscopy, THz sensing have become an important approach of biological and medical detections. Particularly, benefit from the excellent capability of metasurface, strong interactions between the metasurface and THz waves can be realized, thus the THz sensing with high sensitivity becomes reality. However, the common configuration of metasurface-based THz sensor is composed of metallic subwavelength structure. Due to the inherent resistive loss of metal, it is still a great challenge to further enhance the quality factor (Q-factor) of resonance and sensitivity of THz sensor. In this work, we designed an all-dielectric metasurface with high Q-factor for highly sensitive THz sensors. The metasurface is a windmill-like structure consisting of four cuboids, and every adjacent two cuboids are arranged alternately vertically and horizontally. The transmission spectrum of metasurface exhibits four polarization-independent and strong resonance peaks with high Q-factor in 0.1–2.5 THz, and all of them show high sensitivity related to ambient refractive index. The transmitted structure and polarization-independent resonances can relief the difficulty of measurement. We believe these studies will lay the theoretical and technical foundation for the design of high-sensitivity terahertz sensing.

With the rapid advancement of modern technology and radar detection systems, electromagnetic (EM) stealth technology has become increasingly significant, particularly in aircraft stealth and military radar applications. In this work, an all-metal metasurface is designed for broadband terahertz radar cross-section (RCS) reduction and infrared invisibility. The all-metal metasurface possesses extremely low infrared emissivity and high polarization conversion in the terahertz band. Through the joint simulation of MATLAB and CST, a genetic algorithm is used to optimize the random phase distribution of 2, 3, and 4-bit metasurfaces, so that the reflected wave is scattered to achieve broadband terahertz RCS reduction. Simulation results show that the metasurface can simultaneously achieve broadband terahertz RCS reduction in 3–5 THz and infrared invisibility in 24–38 THz (8–12.5 μm). The RCS reduction of the coding metasurface is greater than 10 dB compared to the metal plate, and the maximum RCS reduction of the 4-bit metasurface can reach 21.1 dB. Compared to the traditional design method, the proposed method can reduce time consumption and find the optimal result to achieve high performance. We believe the proposed method can provide significant guidance for surface coating in camouflage applications and opens up new possibilities for improving the information capacity of coding metasurfaces.

Remote sensing image change detection method plays a great role in land cover research, disaster assessment, medical diagnosis, video surveillance, and other fields, so it has attracted wide attention. Based on a small sample dataset from SZTAKI AirChange Benchmark Set, in order to solve the problem that the deep learning network needs a large number of samples, this work first uses nongenerative sample augmentation method and generative sample augmentation method based on deep convolutional generative adversarial networks, and then, constructs a remote sensing image change detection model based on an improved DeepLabv3+ network. This model can realize end-to-end training and prediction of remote sensing image change detection with subpixel convolution. Finally, Landsat 8, Google Earth, and Onera satellite change detection datasets are used to verify the generalization performance of this network. The experimental results show that the improved network accuracy is 95.1% and the generalization performance is acceptable.

Partial Sum Minimization of Singular Values (PSSV) is a powerful tool for image denoising, matrix completion and recovering underlying low-rank structure from the corrupted data via Partial Sum Minimization of Singular Values. However, the performance of PSSV degenerates remarkably when data is incomplete or some data is corrupted completely, which is usually faced in real applications especial for video sequence analysis. To handle this problem, we impose the variance regularization in PSSV by analyzing PSSV and truncated nuclear norm. Our proposed model benefits from 1) the robustness of principal components to outliers and missing values. 2) It can guarantee the compactness of the learned clean data. 3) It can recover the missing sample data from the data matrix. Experimental results on background extraction from the incomplete videos and data, image denoising, and clustering illustrate the effectiveness of the proposed approach.

The acquisition of target data for underwater acoustic target recognition (UATR) is difficult and costly. Although deep neural networks (DNN) have been used in UATR, and some achievements have been made, the performance is not satisfactory when recognizing underwater targets with different Doppler shifts, signal-to-noise ratios (SNR), and interferences. On the basis of this, this paper proposed a Siamese network with two identical one-dimensional convolutional neural networks (1D-CNN) that recognize the detection of envelope modulation on noise (DEMON) spectra of underwater target-radiated noise. The parameters of underwater samples were diverse, but the states of the collected samples were very homogeneous. Traditional underwater target recognition uses multi-state samples to train the network, which is costly. This article trained the network using samples from a single state. The expectation was to be able to identify samples with different parameters. Datasets of targets with different Doppler shifts, SNRs, and interferences were designed to evaluate the generalization performance of the proposed Siamese network. The experimental results showed that when recognizing samples with Doppler shifts, the classification accuracy of the proposed network reached 95.3%. For SNRs, the classification accuracy reached 85.5%. The outstanding generalization ability of the proposed model shows that it is suitable for practical engineering applications.

In the past decades, metasurfaces have shown their extraordinary abilities on manipulating the wavefront of electromagnetic wave. Based on the ability, various kinds of metasurfaces are designed to realize new functional metadevices based on wavefront manipulations, such as anomalous beam steering, focus metalens, vortex beams generator, and holographic imaging. However, most of the previously proposed designs based on metasurfaces are fixed once design, which is limited for applications where light modulation needs to be tunable. In this paper, we proposed a design for THz tunable wavefront manipulation achieved by the combination of plasmonic metasurface and phase change materials (PCMs) in THz region. Here, we designed a metal-insulator-metal (MIM) metasurface with the typical C-shape split ring resonator (CSRR), whose polarization conversion efficiency is nearly 90% for circular polarized light (CPL) in the range of 0.95~1.15 THz when PCM is in the amorphous state, but the conversion efficiency turns to less than 10% in the same frequency range when PCM switches into the crystalline state. Then, benefiting from the high polarization conversion contrast of unit cell, we can achieve tunable wavefront manipulation by utilizing the Pancharatnam–Berry (PB) phase between the amorphous and crystalline states. As a proof-of-concept, the reflective tunable anomalous beam deflector and focusing metalens are designed and characterized, and the results further verify their capability for tunable wavefront manipulation in THz range. It is believed that the design in our work may pave the way toward the tunable wavefront manipulation of THz waves and is potential for dynamic tunable THz devices.

Image registration is an important basis of image processing, which is of great significance in image mosaicking, target recognition, and change detection. Aiming at the automatic registration problem of multi-angle optical images for ground scenes, a registration method combining point features and line features to register images is proposed. Firstly, the LSD (Line Segment Detector) algorithm is used to extract line features of images. The obtained line segments whose length are less than a given threshold are eliminated by a visual significant algorithm. Then, an affine transform model obtained by estimating a Gaussian mixture model (GMM) is applied to the image to be matched. Lastly, Harris point features are utilized in fine matching to overcome shortages of methods based on line features. In experiments, the proposed algorithm is compared with popular feature-based registration algorithms. The results indicate that the proposed algorithm in this work has obvious advantages in terms of registration accuracy and reliability for optical images acquired at different angles.

With the rapid development of unmanned aerial vehicle (UAV) technology, UAV remote sensing images are increasing sharply. However, due to the limitation of the perspective of UAV remote sensing, the UAV images obtained from different viewpoints of a same scene need to be stitched together for further applications. Therefore, an automatic registration method of UAV remote sensing images based on deep residual features is proposed in this work. It needs no additional training and does not depend on image features, such as points, lines and shapes, or on specific image contents. This registration framework is built as follows: Aimed at the problem that most of traditional registration methods only use low-level features for registration, we adopted deep residual neural network features extracted by an excellent deep neural network, ResNet-50. Then, a tensor product was employed to construct feature description vectors through exacted high-level abstract features. At last, the progressive consistency algorithm (PROSAC) was exploited to remove false matches and fit a geometric transform model so as to enhance registration accuracy. The experimental results for different typical scene images with different resolutions acquired by different UAV image sensors indicate that the improved algorithm can achieve higher registration accuracy than a state-of-the-art deep learning registration algorithm and other popular registration algorithms.

This article has been retracted. Please see the Retraction Notice for more detail: https://doi.org/10.1186/s12868-021-00644-1

Water body extraction can help eco-environmental policymakers to intuitively grasp surface water resources. Remote sensing technology can accurately and quickly extract surface water information, which is of great significance for monitoring surface water changes. Fengyun satellite images have the advantages of high time resolution and multispectral bands. This provides important image data suitable for high-frequency surface water monitoring. Based on Fengyun 3 medium resolution spectral imager (FY-3/MERSI) data, 7 methods were applied in this study, which include single-band threshold method, water body index method, knowledge decision tree classification method, supervised classification method, unsupervised classification method, spectral matching based on discrete particle swarm optimization (SMDPSO), and improved spectral matching based on discrete particle swarm optimization with linear feature enhancement (SMDPSO+LFE). These methods were used to extract the land surface water of Poyang Lake, check the samples from the Landsat image with similar times to the FY-3 images, and calculate the classification accuracy via the confusion matrix. The results showed that the overall classification accuracy (OA) of the SMDPSO+LFE is 97.64%, and the Kappa coefficient is 0.95. To analyze the stability of the surface water extracted by SMDPSO+LFE in different regions, this paper selected eight test sites with different surface water types, landscapes, and terrains to extract surface water. Based on an analysis of the land surface water results at the eight test sites, every OA in the eight sites was higher than 94.5%, the Kappa coefficient was greater than 0.88. In conclusion, the SMDPSO+LFE is found to be the most suitable method among the 7 methods and effectively distinguish between different surface water bodies and backgrounds with good stability.

Vehicle targets in unmanned aerial vehicle (UAV) images are generally small, so a significant amount of detailed information on targets may be lost after neural computing, which leads to the poor performances of the existing recognition algorithms. Based on convolutional neural networks that utilize the YOLOv3 algorithm, this article focuses on the development of a quick automatic vehicle detection method for UAV images. First, a vehicle dataset for target recognition is constructed. Then, a novel YOLOv3 vehicle detection framework is proposed according to the following characteristics: The vehicle targets in the UAV image are relatively small and dense. The average precision (AP) increased by 5.48%, from 92.01% to 97.49%, which still remains the rather high processing speed of the YOLO network. Finally, the proposed framework is tested using three datasets: COWC, VEDAI, and CAR. The experimental results demonstrate that our method had a better detection capability.

An accurate inversion of the fraction of absorbed photosynthetically active radiation (FPAR) based on remote sensing data is particularly important for understanding global climate change. At present, there are relatively few studies focusing on the inversion of FPAR using Chinese autonomous satellites. This work intends to investigate the inversion of the FPAR obtained from the FengYun-3C (FY-3C) data of domestic satellites by using the PROSAIL model and the look-up table (LUT) algorithm for different vegetation types from various places in China. After analyzing the applicability of existing models using FY-3C data and MOD09GA data, an inversion strategy for FY-3C data is implemented. This strategy is applied to areas with various types of vegetation, such as grasslands, croplands, shrubs, broadleaf forests, and needleleaf forests, and produces FPAR products, which are cross-validated against the FPAR products from the Moderate Resolution Imaging Spectro Radiometer (MODIS), Geoland Version 1 (GEOV1), and Global Land Surface Satellite (GLASS). Accordingly, the results show that the FPAR retrieved from the FY-3C data has good spatial and temporal consistency and correlation with the three FPAR products. However, this technique does not favor all types of vegetation equally; the FY-FPAR is relatively more suitable for the inversion of grasslands and croplands during the lush period than for others. Therefore, the inversion strategy provides the potential to generate large-area and long-term sequence FPAR products from FY-3C data.