Search

Image–text multimodal deep semantic segmentation leverages the fusion and alignment of image and text information and provides more prior knowledge for segmentation tasks. It is worth exploring image–text multimodal semantic segmentation for remote sensing images. In this paper, we propose a bidirectional feature fusion and enhanced alignment-based multimodal semantic segmentation model (BEMSeg) for remote sensing images. Specifically, BEMSeg first extracts image and text features by image and text encoders, respectively, and then the features are provided for fusion and alignment to obtain complementary multimodal feature representation. Secondly, a bidirectional feature fusion module is proposed, which employs self-attention and cross-attention to adaptively fuse image and text features of different modalities, thus reducing the differences between multimodal features. For multimodal feature alignment, the similarity between the image pixel features and text features is computed to obtain a pixel–text score map. Thirdly, we propose a category-based pixel-level contrastive learning on the score map to reduce the differences among the same category's pixels and increase the differences among the different categories' pixels, thereby enhancing the alignment effect. Additionally, a positive and negative sample selection strategy based on different images is explored during contrastive learning. Averaging pixel values across different training images for each category to set positive and negative samples compares global pixel information while also limiting sample quantity and reducing computational costs. Finally, the fused image features and aligned pixel–text score map are concatenated and fed into the decoder to predict the segmentation results. Experimental results on the ISPRS Potsdam, Vaihingen, and LoveDA datasets demonstrate that BEMSeg is superior to comparison methods on the Potsdam and Vaihingen datasets, with improvements in mIoU ranging from 0.57% to 5.59% and 0.48% to 6.15%, and compared with Transformer-based methods, BEMSeg also performs competitively on LoveDA dataset with improvements in mIoU ranging from 0.37% to 7.14%. [ABSTRACT FROM AUTHOR]

Biochar (BC) application to croplands aims to sequester carbon and improve soil quality, but its impact on soil organic carbon (SOC) dynamics is not represented in most land models used for assessing land-based climate change mitigation; therefore, we are unable to quantify the effects of biochar application under different climate or land management conditions. Here, to fill this gap, we implement a submodel to represent biochar in a microbial decomposition model named MIMICS (MIcrobial-MIneral Carbon Stabilization). We first calibrate and validate MIMICS with new representations of the density-dependent microbial turnover rate, adsorption of available organic carbon on mineral soil particles, and soil moisture effects on decomposition using global field-measured cropland SOC at 285 sites. We further integrate biochar in MIMICS by accounting for its effect on microbial decomposition and SOC sorption/desorption and optimize two biochar-related parameters in these processes using 134 paired SOC measurements with and without biochar addition. The MIMICS-biochar version can generally reproduce the short-term (≤ 6 years) and long-term (8 years) SOC changes after adding (mean addition rate of 25.6 t ha -1) biochar (R2= 0.79 and 0.97, respectively) with a low root-mean-square error (RMSE = 3.73 and 6.08 g kg -1 , respectively). Our study incorporates sorption and soil moisture processes into MIMICS and extends its capacity to simulate biochar decomposition, providing a useful tool to couple with dynamic land models to evaluate the effectiveness of biochar application with respect to removing CO 2 from the atmosphere. [ABSTRACT FROM AUTHOR]

Phytoremediation is considered an effective strategy for remediation of heavy-metal-contaminated soil in mining areas. However, single-species plants cannot reach the highest potential for uptake of heavy metals due to inhibition of their growth by high concentrations of heavy metals in the soil. Therefore, this study has explored the effects of illite application and two plant species' co-cropping on soil quality, plant growth, and heavy metal transformation in a soil–plant system. The results reveal that the addition of 1% (mass fraction) of illite significantly enhances soil pH. The co-cropping of Ilex cornuta and Epipremnum aureum is beneficial for improving the organic matter content of the soil. The contents of EDTA-extractable Pb, Zn, and Cu were significantly reduced by 29.8–32.5%, 1.85–5.72%, and 30.0–32.9%, respectively, compared to the control. The co-cropping of Ilex cornuta and Epipremnum aureum promoted enrichment effects of Epipremnum aureum on Pb and Ilex cornuta on Cd (p < 0.05). The co-cropping pattern lowered the biomass of Ilex cornuta and Epipremnum aureum; however, co-cropping of Ilex cornuta and Epipremnum aureum promoted the elimination of Pb, Zn, Cu, and Cd from the soil at 13.0–75.8%, 11.1–38.2%, 8.39–88.4%, and 27.8–72.5%, respectively. It is concluded that illite application combined with co-cropping of Ilex cornuta and Epipremnum aureum is highly effective for the elimination of Pb, Zn, Cu, and Cd from contaminated soil. This study provides a theoretical basis and pathway for the restoration of heavy-metal-contaminated soil in mining with the application of bentonite combined with phytoremediation. [ABSTRACT FROM AUTHOR]

Pantograph-catenary system is the key part of train current collection. The current density distribution of current-carrying friction pair used in pantograph-catenary system affects the safety running of train. At present, there is little research on the current density distribution of the friction pair. Because it is difficult to measure the current density directly, it is necessary to study by simulation. A sliding electrical contact simulation model of pantograph-catenary system was established by COMSOL Multiphysics software. The current density distribution of the friction pair under normal, light, and heavy rain conditions were studied with the simulation model. The current density isosurface in contact wire is saddle-shaped, and that in slide is arc-shaped. Under normal condition, the current density of contact area increases firstly and then decreases with the increase of contact force. It increases with the increase of current and decreases with the increase of sliding speed. Under rain condition, the current density decreases firstly and then increases with the increase of contact force. It increases with the increase of current. It increases firstly, then decreases and finally increases with the increase of sliding speed. Under the same contact force, sliding speed, and current conditions, the current density under rain condition is lower than that under no rain condition. Finally, the validity of the model was verified.

In order to investigate the wear characteristics of pantograph slide in pantogragh-catenary under different current, contact pressure and sliding speed, current-carrying wear experiments were carried out to study the current-carrying efficiency, arc energy, friction coefficient and temperature of the friction pair under different working conditions. The wear area of the slide was measured in blocks and multiple times after the experiments, and the surface topography images and roughness parameters of the slides were collected by metallographic microscope and surface roughness instrument respectively. An image processing method based on Otsu algorithm was used to extract pits from the topography images and perform mark calculation, and the variation of the area and number of pits, and the roughness parameters (contour arithmetic mean deviation and maximum contour valley depth) under different working conditions was statistically analyzed. The results show that with the increase of current, the area and number of pits increase, and the roughness parameters decrease. With the increase of contact pressure, the area of pits decreases, the number of pits decreases first and then increases slightly, and the roughness parameters decrease. With the increase of sliding speed, the area and number of pits and the roughness parameters increase. [ABSTRACT FROM AUTHOR]

Hyperspectral images (HSIs) contain hundreds of continuous spectral bands and are rich in spectral‐spatial information. In terms of HSIs' classification, traditional convolutional neural networks (CNNs) extract features based on HSI's spectral‐spatial information through 2D convolution. However, 2D convolution extracts features in 2D plane without considering the relationships between spectral bands, which inevitably leads to insufficient feature extraction. 3D convolutional neural networks (3DCNNs) take account of the correlations among spectral bands and outperform 2D convolutional networks in feature extraction, but the computational cost is rather expensive. To address the above problem, a light‐weight three‐layer 3D convolutional network Module (3D‐M) for HSIs' spectral‐spatial feature extraction is proposed. Another challenge is that neither 2D convolution nor 3D convolution utilizes the structural information inherent in the data. Graph convolution networks (GCNs) can model and utilize such information through the similarity matrix, also known as adjacency matrix. However, traditional GCNs cannot handle large‐scale data because they construct adjacency matrix on all data, which results in high computational complexity and large storage requirement. To conquer this challenge, this article proposes a batch‐graph strategy on which a scalable GCN is developed. Finally, a hybrid network model (HNM) based on the proposed light‐weight 3D‐M and scalable GCN is presented. HNM extracts spectral‐spatial features of HSIs with low computational complexity through the light‐weight 3D convolution network and leverages the structural information in data via the scalable GCN. The experimental results on three public datasets with different sizes demonstrate that the proposed HNM produces better classification results than other state‐of‐the‐art hyperspectral images classification models in terms of overall accuracy (OA), average accuracy (AA) and kappa coefficient (Kappa). [ABSTRACT FROM AUTHOR]