Search Results (75)

Panoptic scene graph generation (PSG) aims to simultaneously segment both foreground objects and background regions while predicting object relations for fine-grained scene modeling. Despite significant progress in panoptic scene understanding, current PSG methods face challenging problems: relation prediction often only relies on visual representations and is hindered by imbalanced relation category distributions. Accordingly, we propose IntegraPSG, a single-stage framework that integrates large language model (LLM) guidance with multimodal feature fusion. IntegraPSG introduces a multimodal sparse relation prediction network that efficiently integrates visual, linguistic, and depth cues to identify subject–object pairs most likely to form relations, enhancing the screening of subject–object pairs and filtering dense candidates into sparse, effective pairs. To alleviate the long-tail distribution problem of relations, we design a language-guided multimodal relation decoder where LLM is utilized to generate language descriptions for relation triplets, which are cross-modally attended with vision pair features. This design enables more accurate relation predictions for sparse subject–object pairs and effectively improves discriminative capability for rare relations. Experimental results show that IntegraPSG achieves steady and strong performance on the PSG dataset, especially with the R@100, mR@100, and mean reaching 38.7%, 28.6%, and 30.0%, respectively, indicating strong overall results and supporting the validity of the proposed method. Full article

Phosphorite, or phosphate rock, has garnered increasing attention in recent years as a promising unconventional resource for rare earth elements (REEs). This paper presents a processing scheme aimed at recovering both REEs and phosphate values from amine flotation tailings generated during phosphate beneficiation in Florida. In these tailings, REEs are primarily present as monazite and xenotime, often associated with heavy minerals. The proposed flowsheet includes gravity separation to pre-concentrate REE- and phosphate-bearing minerals, followed by flotation to further upgrade both REEs and phosphate, and finally sulfuric acid leaching to extract REEs and phosphate from the flotation concentrate. Gravity separation using a shaking table increased the total REE content from approximately 202 ppm to 657 ppm, with a concentrate yield of 12.51%, REE recovery of around 41%, and P₂O₅ recovery of 33%. Fatty acid flotation of the shaking table concentrate produced a final concentrate containing 1106 ppm REEs and 14.90% P₂O₅, with recoveries of approximately 86% for REEs and 90% for P₂O₅. Subsequent pyrolysis with concentrated sulfuric acid followed by water leaching achieved recoveries of about 85% for REEs and 93% for P₂O₅. While the process demonstrated effective concentration and leaching of REE minerals and apatite, the major challenge to further improving separation and extraction efficiency lies in the fine-grained nature of the valuable minerals and their interlocking with gangue minerals. Full article

Phonolite is a fine-grained, shallow extrusive rock rich in alkali minerals and containing iron/titanium-bearing minerals. This rock is widely used as a construction material for building exteriors due to its excellent abrasion resistance and insulation properties. However, during the cutting process, approximately 70% of the rock is discarded as tailing. So, this study aims to repurpose tailings from a phonolite cutting and sizing plant into a high-alkali ceramic raw mineral concentrate. To enable the use of phonolite tailings in ceramic manufacturing, it is necessary to remove coloring iron/titanium-bearing minerals, which negatively affect the final product. To achieve this removal, dry/wet magnetic separation processes, along with flotation, were employed both individually and in combination. The results demonstrated that using dry high-intensity magnetic separation (DHIMS) resulted in a concentrate with an Fe₂O₃ + TiO₂ grade of 0.95% and a removal efficiency of 85%. The wet high-intensity magnetic separation (WHIMS) process reduced the Fe₂O₃ + TiO₂ grade of the concentrate to 1.2%, with 70% removal efficiency. During flotation tests, both pH levels and collector concentration impacted the efficiency and Fe₂O₃ + TiO₂ grade (%) of the concentrate. The lowest Fe₂O₃ + TiO₂ grade of 1.65% was achieved at a pH level of 10 with a collector concentration of 2000 g/t. Flotation concentrates processed with DHIMS achieved a minimum Fe₂O₃ + TiO₂ grade of 0.90%, while those processed with WHIMS exhibited higher Fe₂O₃ + TiO₂ grades (>1.1%) and higher recovery rates (80%). Additionally, studies on flotation applied to WHIMS concentrates showed that collector concentration, pulp density, and conditioning time significantly influenced the Fe₂O₃ + TiO₂ grade of the final concentrate. Full article

The overtopping-induced lateral erosion breaching of tailings dams represents a critical disaster mechanism threatening structural safety, particularly in reinforced fine-grained tailings dams where erosion behaviors demonstrate pronounced water–soil coupling characteristics and material anisotropy. Through physical model tests and numerical simulations, this study systematically investigates lateral erosion failure patterns of reinforced fine-grained tailings under overtopping flow conditions. Utilizing a self-developed hydraulic initiation test apparatus, with aperture sizes of reinforced geogrids (2–3 mm) and flow rates (4–16 cm/s) as key control variables, the research elucidates the interaction mechanisms of “hydraulic scouring-particle migration-geogrid anti-sliding” during lateral erosion processes. The study revealed that compared to unreinforced specimens, reinforced specimens with varying aperture sizes (2–3 mm) demonstrated systematic reductions in final lateral erosion depths across flow rates (4–16 cm/s): 3.3–5.8 mm (15.6−27.4% reduction), 3.1–7.2 mm (12.8–29.6% reduction), 2.3–11 mm (6.9–32.8% reduction), and 2.5–11.4 mm (6.2–28.2% reduction). Smaller-aperture geogrids (2 mm × 2 mm) significantly enhanced anti-erosion performance through superior particle migration inhibition. Concurrently, a pronounced positive correlation between flow rate and lateral erosion depth was confirmed, where increased flow rates weakened particle erosion resistance and exacerbated lateral erosion severity. The numerical simulation results are in basic agreement with the lateral erosion failure process observed in model tests, revealing the dynamic process of lateral erosion in the overtopping breach of a reinforced tailings dam. These findings provide critical theoretical foundations for optimizing reinforced tailings dam design, construction quality control, and operational maintenance, while offering substantial engineering applications for advancing green mine construction. Full article

Due to the challenges posed by high resolution, substantial background noise, significant object scale variation, and long-tailed data distribution in remote sensing images, traditional techniques often struggle to maintain both high accuracy and low latency. This paper proposes YOLO11-FSDAT, an advanced object detection framework tailored for remote sensing imagery, which integrates not only modular enhancements but also theoretical and architectural innovations to address these limitations. First, we propose the frequency–spatial feature extraction fusion module (Freq-SpaFEFM), which breaks the conventional paradigm of spatial-domain-dominated feature learning by introducing a multi-branch architecture that fuses frequency- and spatial-domain features in parallel. This design provides a new processing paradigm for multi-scale object detection, particularly enhancing the model’s capability in handling dense and small-object scenarios with complex backgrounds. Second, we introduce the deformable attention-based global–local fusion module (DAGLF), which combines fine-grained local features with global context through deformable attention and residual connections. This enables the model to adaptively capture irregularly oriented objects (e.g., tilted aircraft) and effectively mitigates the issue of information dilution in deep networks. Third, we develop the adaptive threshold focal loss (ATFL), which is the first loss function to systematically address the long-tailed distribution in remote sensing datasets by dynamically adjusting focus based on sample difficulty. Unlike traditional focal loss with fixed hyperparameters, ATFL decouples hard and easy samples and automatically adapts to varying class distributions. Experimental results on the public DOTAv1, SIMD, and DIOR datasets demonstrated that YOLO11-FSDAT achieved 75.22%, 82.79%, and 88.01% mAP, respectively, outperforming baseline YOLOv11n by up to 4.11%. These results confirm the effectiveness, robustness, and broader theoretical value of the proposed framework in addressing key challenges in remote sensing object detection. Full article

With the rapid development of remote sensing technology, satellites can now capture multispectral (MS) and panchromatic (PAN) images simultaneously. MS images offer rich spectral details, while PAN images provide high spatial resolutions. Effectively leveraging their complementary strengths and addressing modality gaps are key challenges in improving the classification performance. From the perspective of deep learning, this paper proposes a novel dual-source remote sensing classification framework named the Diversity Extraction and Fusion Classifier (DEFC-Net). A central innovation of our method lies in introducing a modality-specific intra-class diversity modeling mechanism for the first time in dual-source classification. Specifically, the intra-class diversity identification and splitting (IDIS) module independently analyzes the intra-class variance within each modality to identify semantically broad classes, and it applies an optimized K-means method to split such classes into fine-grained sub-classes. In particular, due to the inherent representation differences between the MS and PAN modalities, the same class may be split differently in each modality, allowing modality-aware class refinement that better captures fine-grained discriminative features in dual perspectives. To handle the class imbalance introduced by both natural long-tailed distributions and class splitting, we design a long-tailed ensemble learning module (LELM) based on a multi-expert structure to reduce bias toward head classes. Furthermore, a dual-modal knowledge distillation (DKD) module is developed to align cross-modal feature spaces and reconcile the label inconsistency arising from modality-specific class splitting, thereby facilitating effective information fusion across modalities. Extensive experiments on datasets show that our method significantly improves the classification performance. The code was accessed on 11 April 2025. Full article

This study develops a comprehensive, data-driven framework for predicting satellite imagery prices using four state-of-the-art ensemble learning algorithms: XGBoost, LightGBM, AdaBoost, and CatBoost. Two distinct datasets—optical and Synthetic Aperture Radar (SAR) imagery—were assembled, each characterized by nine technical and economic features (e.g., imaging mode, spatial resolution, satellite manufacturing cost, and acquisition timeliness). Bayesian optimization is employed to systematically tune hyperparameters, thereby minimizing overfitting and maximizing generalization. Models are evaluated on held-out test sets (20% of data) using Pearson’s correlation coefficient (R), mean bias error (MBE), root mean square error (RMSE), unbiased RMSE (ubRMSE), Nash–Sutcliffe Efficiency (NSE), and Kling–Gupta Efficiency (KGE). For optical imagery, the Bayesian-optimized XGBoost model achieves the best performance ($R=0.9870$, $\mathrm{RMSE}=\$3.44/{\mathrm{km}}^2$, $\mathrm{NSE}=0.9651$, $\mathrm{KGE}=0.8950$), followed closely by CatBoost ($R=0.9826$, $\mathrm{RMSE}=\$3.83/{\mathrm{km}}^2$). For SAR imagery, CatBoost outperforms all others after optimization ($R=0.9278$, $\mathrm{RMSE}=\$9.94/{\mathrm{km}}^2$, $\mathrm{NSE}=0.8575$, $\mathrm{KGE}=0.8443$), reflecting its robustness to heavy-tailed price distributions. AdaBoost also demonstrates competitive accuracy, while LightGBM and XGBoost exhibit larger errors in high-value regimes. SHapley Additive exPlanations (SHAP) analysis reveals that imaging mode and spatial resolution are the primary drivers of price variance across both domains, followed by satellite manufacturing cost and acquisition recency. These insights demonstrate how ensemble models capture nonlinear, high-dimensional interactions that traditional rule-based pricing schemes overlook. Compared to static, experience-driven price brackets, our machine learning approach provides a scalable, transparent, and economically rational pricing engine—adaptable to rapidly changing market conditions and capable of supporting fine-grained, application-specific pricing strategies. Full article

Enhancing the detection capabilities of marine vessels is crucial for maritime security and intelligence acquisition. However, accurately identifying small ships in complex oceanic environments remains a significant challenge, as these targets are frequently obscured by ocean waves and other disturbances, compromising recognition accuracy and stability. To address this issue, we propose YOLO-ssboat, a novel small-target ship recognition algorithm based on the YOLOv8 framework. YOLO-ssboat integrates the C2f_DCNv3 module to extract fine-grained features of small vessels while mitigating background interference and preserving critical target details. Additionally, it employs a high-resolution feature layer and incorporates a Multi-Scale Weighted Pyramid Network (MSWPN) to enhance feature diversity. The algorithm further leverages an improved multi-attention detection head, Dyhead_v3, to refine the representation of small-target features. To tackle the challenge of wake waves from moving ships obscuring small targets, we introduce a gradient flow mechanism that improves detection efficiency under dynamic conditions. The Tail Wave Detection Method synergistically integrates gradient computation with target detection techniques. Furthermore, adversarial training enhances the network’s robustness and ensures greater stability. Experimental evaluations on the Ship_detection and Vessel datasets demonstrate that YOLO-ssboat outperforms state-of-the-art detection algorithms in both accuracy and stability. Notably, the gradient flow mechanism enriches target feature extraction for moving vessels, thereby improving detection accuracy in wake-disturbed scenarios, while adversarial training further fortifies model resilience. These advancements offer significant implications for the long-range monitoring and detection of maritime vessels, contributing to enhanced situational awareness in expansive oceanic environments. Full article

This study developed a novel pulsation-fluidized bed system, and the device was integrated into a reflux classifier to enhance the preconcentration of antimony oxide ore. The diaphragm-based pulsation device converts a stable upward water flow into a vertically alternating pulsation flow. By precisely controlling the pulsation parameters and optimizing operational conditions, the density-based stratification of particles can be significantly enhanced, thereby improving bed layering and effectively reducing entrainment. An antimony oxide ore from flotation tailings with an Sb grade of 0.8% was used as the feed material to evaluate the performance of the pulsation reflux classifier (PRC). Under optimized conditions, the PRC produced a concentrate with an Sb grade of 5.48% and a recovery of 81.68%, corresponding to a high separation efficiency of 70.97%. The response surface statistical model revealed that the interaction between the fluidization rate and pulsation frequency significantly enhanced the Sb grade of the concentrate, while pulsation stroke was identified as the key factor influencing separation efficiency. Furthermore, the variation in bed profile parameters with changing pulsation characteristics elucidates the interplay between particle suspension, stratification, and fluid disturbances. This study demonstrates that pulsation fluidization significantly enhances the separation performance of the reflux classifier, offering a new approach for the efficient preconcentration of complex fine-grained minerals. Full article

Facial semantic segmentation, as a critical technology in high-level visual understanding, plays an important role in applications such as facial editing, augmented reality, and identity recognition. However, due to the complexity of facial structures, ambiguous boundaries, and inconsistent scales of facial components, traditional methods still suffer from significant limitations in detail preservation and contextual modeling. To address these challenges, this paper proposes a facial parsing network based on the Deeplabv3+ framework, named FP-Deeplab, which aims to improve segmentation performance and generalization capability through structurally enhanced modules. Specifically, two key modules are designed: (1) the Context-Channel Refine Feature Enhancement (CCR-FE) module, which integrates multi-scale contextual strip convolutions and Cross-Axis Attention and introduces a channel attention mechanism to strengthen the modeling of long-range spatial dependencies and enhances the perception and representation of boundary regions; (2) the Self-Modulation Attention Feature Integration with Regularization (SimFA) module, which combines local detail modeling and a parameter-free channel attention modulation mechanism to achieve fine-grained reconstruction and enhancement of semantic features, effectively mitigating boundary blur and information loss during the upsampling stage. The experimental results on two public facial segmentation datasets, CelebAMask-HQ and HELEN, demonstrate that FP-Deeplab improves the baseline model by 3.8% in Mean IoU and 2.3% in the overall F1-score on the HELEN dataset, and it achieves a Mean F1-score of 84.8% on the CelebAMask-HQ dataset. Furthermore, the proposed method shows superior accuracy and robustness in multiple key component categories, especially in long-tailed regions, validating its effectiveness. Full article

Class imbalance (CI) poses a significant challenge in machine learning, characterized by a substantial disparity in sample sizes between majority and minority classes, leading to a pronounced “long-tail effect” in statistical distributions and subsequent inference processes. This issue is particularly acute in high-resolution land cover classification within arid regions, where CI tends to bias classification outcomes towards majority classes, often at the expense of minority classes. Recent advancements in deep learning have opened new avenues for tackling the CI problem in this context, focusing on three key aspects: the semantic segmentation model, loss function design, and dataset composition. To address this issue, we propose the high-resolution U-shaped mamba network (HRUMamba), which integrates multiple innovations to enhance segmentation performance under imbalanced conditions. Specifically, HRUMamba adopts a pre-trained HRNet as the encoder for capturing fine-grained local features and incorporates a modified scaled visual state space (SVSS) block in the decoder to model long-range dependencies effectively. An adaptive awareness fusion (AAF) module is embedded within the skip connections to enhance target saliency. Additionally, we introduce a synthetic loss function that combines cross-entropy loss, Dice loss, and auxiliary loss to improve optimization stability. To quantitatively assess multi-class imbalance, we introduce the coefficient of variation (CV) as a novel evaluation metric. Experimental results on the ISPRS Vaihingen and Minqin datasets demonstrate the robustness and effectiveness of HRUMamba in mitigating CI. The proposed model achieves the highest mF1 scores of 92.25% and 89.88%, along with the lowest CV values of 0.0445 and 0.0574, respectively, outperforming state-of-the-art methods. These innovations underscore the potential of HRUMamba in advancing high-resolution land cover classification in imbalanced datasets. Full article

The characterization of historical tailings bodies is crucial for optimizing environmental management and resource recovery efforts. This study investigated the Bielatal tailings dam (Altenberg, Germany), examining its internal structure, material distribution influenced by historical flushing technology, and the spatial distribution of valuable elements. To evaluate the tailings resource potential, drill core sampling was conducted at multiple points at a depth of 7 m. Subsequent analyses included geochemical characterization using sodium peroxide fusion, lithium borate fusion, X-ray fluorescence (XRF), and a scanning electron microscope with energy dispersive X-ray spectroscopy (SEM-EDX). Particle size distribution analysis via a laser particle size analyzer and wet sieving was conducted alongside milieu parameter (pH, Eh, EC) analysis. A theoretical assessment of the tailings’ potential for geopolymer applications was conducted by comparing them with other tailings used in geopolymer research and relevant European standards. The results indicated average concentrations of lithium (Li) of 0.1 wt%, primarily hosted in Li-mica phases, and concentrations of tin (Sn) of 0.12 wt%, predominantly occurring in cassiterite. Particle size analysis revealed that the tailings material is generally fine-grained, comprising approximately 60% silt, 32% fine sand, and 8% clay. These textural characteristics influenced the spatial distribution of elements, with Li and Sn enriched in fine-grained fractions predominantly concentrated in the dam’s central and western sections, while coarser material accumulated near injection points. Historical advancements in mineral processing, particularly flotation, had significantly influenced Sn distribution, with deeper layers showing higher Sn enrichment, except for the final operational years, which also exhibited elevated Sn concentrations. Due to the limitations of X-ray fluorescence (XRF) in detecting Li, a strong correlation between rubidium (Rb) and Li was established, allowing Li quantification via Rb measurements across varying particle sizes, redox conditions, and geological settings. This demonstrated that Rb can serve as a reliable proxy for Li quantification in diverse contexts. Geochemical and mineralogical analyses revealed a composition dominated by quartz, mica, topaz, and alkali feldspars. The weakly acidic to neutral conditions (pH 5.9–7.7) and reducing redox potential (Eh, 570 to 45 mV) of the tailings material indicated a minimal risk of acid mine drainage. Preliminary investigations into using Altenberg tailings as geopolymer materials suggested that their silicon-rich composition could serve as a substitute for coal fly ash in construction; however, pre-treatment would be needed to enhance reactivity. This study underscores the dual potential of tailings for element recovery and sustainable construction, emphasizing the importance of understanding historical processing techniques for informed resource utilization. Full article

Typical greisen-type ore samples from northeastern Tasmania were investigated for their critical metal potential. The samples contain zinnwaldite (KLiFe²⁺Al(AlSi₃O₁₀)(F,OH)₂), a lithium-bearing mica that is prone to excessive breakage during conventional processing, leading to the generation of very-fine-sized particles (i.e., slimes, <20 µm), eventually ending up in tailings and resulting in lithium (Li) loss. To assess whether the natural grain size of valuable minerals could be preserved, the samples were processed using electric pulse fragmentation (EPF). The results indicate that EPF preferentially fragmented along mica-rich veins, maintaining coarse grain sizes, although a lower degree of liberation was observed in fine-grained, massive samples. In addition, the critical metal distribution within zinnwaldite was examined using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) techniques. The results reveal differences in Li content between groundmass zinnwaldite and vein-hosted zinnwaldite and that the zinnwaldite contains the critical elements rubidium (Rb), cesium (Cs), and rare earth elements (REEs: La, Ce, Pr, and Nd). Vein-hosted zinnwaldite has a higher average Li content, whereas groundmass mica contains higher concentrations of Rb, Cs, and REEs. Both mica types host inclusions of bismuth–copper–thorium–arsenic (Bi-Cu-Th-As), which are more abundant in vein-hosted mica. In some of the samples, Bi, Cu, Th, and REEs also occur along the mica cleavage planes, as well as in mineral inclusions. The Li, Rb, and Cs grades are comparable to those of European deposits, such as Cínovec and the Zinnwald Lithium Project. Full article

Object detection in unmanned aerial vehicle (UAV) images poses significant challenges due to complex scale variations and class imbalance among objects. Existing methods often address these challenges separately, overlooking the intricate nature of UAV images and the potential synergy between them. In response, this paper proposes AD-Det, a novel framework employing a coherent coarse-to-fine strategy that seamlessly integrates two pivotal components: adaptive small object enhancement (ASOE) and dynamic class-balanced copy–paste (DCC). ASOE utilizes a high-resolution feature map to identify and cluster regions containing small objects. These regions are subsequently enlarged and processed by a fine-grained detector. On the other hand, DCC conducts object-level resampling by dynamically pasting tail classes around the cluster centers obtained by ASOE, maintaining a dynamic memory bank for each tail class. This approach enables AD-Det to not only extract regions with small objects for precise detection but also dynamically perform reasonable resampling for tail-class objects. Consequently, AD-Det enhances the overall detection performance by addressing the challenges of scale variations and class imbalance in UAV images through a synergistic and adaptive framework. We extensively evaluate our approach on two public datasets, i.e., VisDrone and UAVDT, and demonstrate that AD-Det significantly outperforms existing competitive alternatives. Notably, AD-Det achieves a $37.5\%$ average precision (AP) on the VisDrone dataset, surpassing its counterparts by at least $3.1\%$. Full article

Screening is a key step in the mineral process of ilmenite. As the grading particle size decreases, the phenomenon of clogged holes on the screening-plate intensifies, the screening environment deteriorates, and the screening effect deteriorates, seriously restricting subsequent sorting operations. This study proposes a 1 mm dry flip-flow screening method for ilmenite to achieve efficient deep classification of fine-grained materials. Firstly, a laser displacement testing system is used to study the dynamic characteristics of the flip-flow screen; based on the characteristics of different feed particle sizes, further research is conducted on the 1 mm dry flip-flow screening effect of ilmenite under different ratios of obstructive particles and difficult-to-screen particles. The 1 mm screening effect can reach 85.41%. Finally, the pre-throwing tailings process based on 1 mm multi-stage screening is put forward. This pre-throwing tail process has the characteristic of not using water and is suitable for sorting in arid, water deficient, and high-altitude frozen soil areas. It has important promotion and application value. Full article