Search Results (113)

Three-dimensional fragment assembly technology has significant application value in fields such as cultural relic restoration, medical image analysis, and industrial quality inspection. To address the common challenges of limited feature representation ability and insufficient assembling accuracy in existing methods, this paper proposes a geometry-aware hierarchical fragment assembly framework (GeoAssemble). The core contributions of our work are threefold: first, the framework utilizes DGCNN to extract local geometric features while integrating centroid relative positions to construct a multi-dimensional feature representation, thereby enhancing the identification quality of fracture points; secondly, it designs a two-stage matching strategy that combines global shape similarity coarse matching with local geometric affinity fine matching to effectively reduce matching ambiguity; finally, we propose an auxiliary transformation estimation mechanism based on the geometric center of fracture point clouds to robustly initialize pose parameters, thereby improving both alignment accuracy and convergence stability. Experiments conducted on both synthetic and real-world fragment datasets demonstrate that this method significantly outperforms baseline methods in matching accuracy and exhibits higher robustness in multi-fragment scenarios. Full article

In surgical navigation-assisted pedicle screw fixation, cross-source pre- and intra-operative point clouds registration faces challenges like significant initial pose differences and low overlapping ratio. Classical algorithms based on feature descriptor have high computational complexity and are less robust to noise, leading to a decrease in accuracy and navigation performance. To address these problems, this paper proposes a coarse-to-fine registration framework. In the coarse registration stage, a Point Matching algorithm based on Curvature Feature Learning (CFL-PM) is proposed. Through CFL-PM and Farthest Point Sampling (FPS), the coarse registration of overlapping regions between the two point clouds is achieved. In the fine registration stage, the Iterative Closest Point (ICP) is used for further optimization. The proposed method effectively addresses the challenges of noise, initial pose and low overlapping ratio. In noise-free point cloud registration experiments, the average rotation and translation errors reached 0.34° and 0.27 mm. Under noisy conditions, the average rotation error of the coarse registration is 7.28°, and the average translation error is 9.08 mm. Experiments on pre- and intra-operative point cloud datasets demonstrate the proposed algorithm outperforms the compared algorithms in registration accuracy, speed, and robustness. Therefore, the proposed method can achieve the precise alignment of the surgical navigation-assisted pedicle screw fixation. Full article

As the world is technologically advancing, the integration of FSO communication in non-terrestrial platforms is transforming the landscape of global connectivity. By enabling high-data-rate inter-satellite links, secure UAV–ground channels, and efficient HAPS backhaul, FSO technology is paving the way for sustainable 6G non-terrestrial networks. However, the stringent requirement for precise line-of-sight (LoS) alignment between the optical transmitter and receivers poses a hindrance in practical deployment. As non-terrestrial missions require continuous movement across the mission area, the platform is subject to vibrations, dynamic motion, and environmental disturbances. This makes maintaining the LoS between the transceivers difficult. While fine-pointing mechanisms such as fast steering mirrors and adaptive optics are effective for microradian angular corrections, they rely heavily on an initial coarse alignment to maintain the LoS. Coarse pointing modules or gimbals serve as the primary mechanical interface for steering and stabilizing the optical beam over wide angular ranges. This survey presents a comprehensive analysis of coarse pointing and gimbal modules that are being used in FSO communication systems for non-terrestrial platforms. The paper classifies gimbal architectures based on actuation type, degrees of freedom, and stabilization strategies. Key design trade-offs are examined, including angular precision, mechanical inertia, bandwidth, and power consumption, which directly impact system responsiveness and tracking accuracy. This paper also highlights emerging trends such as AI-driven pointing prediction and lightweight gimbal design for SWap-constrained platforms. The final part of the paper discusses open challenges and research directions in developing scalable and resilient coarse pointing systems for aerial FSO networks. Full article

Weakly Supervised Video Anomaly Detection (WSVAD) is a critical task in computer vision. It aims to localize and recognize abnormal behaviors using only video-level labels. Without frame-level annotations, it becomes significantly challenging to model temporal dependencies. Given the diversity of abnormal events, it is also difficult to model semantic representations. Recently, the cross-modal pre-trained model Contrastive Language-Image Pretraining (CLIP) has shown a strong ability to align visual and textual information. This provides new opportunities for video anomaly detection. Inspired by CLIP, WSVAD-CLIP is proposed as a framework that uses its cross-modal knowledge to bridge the semantic gap between text and vision. First, the Axial-Graph (AG) Module is introduced. It combines an Axial Transformer and Lite Graph Attention Networks (LiteGAT) to capture global temporal structures and local abnormal correlations. Second, a Text Prompt mechanism is designed. It fuses a learnable prompt with a knowledge-enhanced prompt to improve the semantic expressiveness of category embeddings. Third, the Abnormal Visual-Guided Text Prompt (AVGTP) mechanism is proposed to aggregate anomalous visual context for adaptively refining textual representations. Extensive experiments on UCF-Crime and XD-Violence datasets show that WSVAD-CLIP notably outperforms existing methods in coarse-grained anomaly detection. It also achieves superior performance in fine-grained anomaly recognition tasks, validating its effectiveness and generalizability. Full article

Multi-modal image segmentation is a key task in various fields such as urban planning, infrastructure monitoring, and environmental analysis. However, it remains challenging due to complex scenes, varying object scales, and the integration of heterogeneous data sources (such as RGB, depth maps, and infrared). To address these challenges, we proposed a novel multi-modal segmentation framework, DyFuseNet, which features dynamic adaptive windows and cross-scale feature fusion capabilities. This framework consists of three key components: (1) Dynamic Window Module (DWM), which uses dynamic partitioning and continuous position bias to adaptively adjust window sizes, thereby improving the representation of irregular and fine-grained objects; (2) Scale Context Attention (SCA), a hierarchical mechanism that associates local details with global semantics in a coarse-to-fine manner, enhancing segmentation accuracy in low-texture or occluded regions; and (3) Hierarchical Adaptive Fusion Architecture (HAFA), which aligns and fuses features from multiple modalities through shallow synchronization and deep channel attention, effectively balancing complementarity and redundancy. Evaluated on benchmark datasets (such as ISPRS Vaihingen and Potsdam), DyFuseNet achieved state-of-the-art performance, with mean Intersection over Union (mIoU) scores of 80.40% and 80.85%, surpassing MFTransNet by 1.91% and 1.77%, respectively. The model also demonstrated strong robustness in challenging scenes (such as building edges and shadowed objects), achieving an average F1 score of 85% while maintaining high efficiency (26.19 GFLOPs, 30.09 FPS), making it suitable for real-time deployment. This work presents a practical, versatile, and computationally efficient solution for multi-modal image analysis, with potential applications beyond remote sensing, including smart monitoring, industrial inspection, and multi-source data fusion tasks. Full article

Planosols are seasonally waterlogged soils characterized by an abrupt transition from coarse-textured surface horizons to dense, clay-enriched subsoils. Despite the increased agricultural expansion in the Planosol landscapes, these soils have been largely overlooked in Ethiopia. The FAO soil map of Ethiopia (1:200,000 scale) does not recognize the presence of Planosols. In contrast, the more recent digital soil map of Ethiopia, EthoSoilGrids v1.0, at a 250 spatial resolution, was not detailed enough to capture Planosol landscapes, reflecting their historical undersampling in the legacy data. To address this gap, we conducted a thorough mapping and characterization of Planosols in the Omo-Gibe basin, southwestern Ethiopian highlands. Using over 200 auger observations, 74 georeferenced soil profiles, 296 laboratory analyses, and Random Forest modeling, we produced a 30 m-resolution soil-landscape map. Our results show that Planosols cover about 18% of the basin, a substantial extent previously unrecognized in national exploratory maps. Morphologically, these soils exhibit abrupt textural change from the coarse-textured, light grey Ap/Eg horizon (about 30–40 cm thick) to a very clayey, grey–black Bssg/Bt horizon occurring below 40 cm depth. Analytical data on selected parameters show the following pattern: low clay contents (20–29%) and acidic pH (5.2–5.8) with relatively low CEC values (11–26 cmol/kg) in the surface horizons (Ap/Eg), but pronounced clay increase (37–74%), higher bulk density (1.3 g/cm³), higher pH (up to 6.5), and substantially higher CEC (37–47 cmol/kg) in the sub-surface horizons (Bss/Bt). In terms of soil fertility, Planosols are low in SOC, TN, and exchangeable K contents, but micronutrient levels are variable—high in Fe-Mn-Zn and low in B and Cu. The findings confirm the diagnostic features of WRB Planosols and align with regional East African averages, underscoring the reproducibility of our approach. By rectifying long-standing misclassifications and generating fine-scale, field-validated evidence on soil fertility constraints and management options, this study establishes a strong foundation for targeted soil management in Ethiopia. It offers transferable insights for Planosol-dominated agroecosystems across Eastern Africa. Globally, the dataset contributes to enriching the global scientific knowledge and evidence base on Planosols, thereby supporting their improved characterization and management. Full article

To tackle the challenges of complex construction environment interference (e.g., lighting variations, occlusion, and marker contamination) and the demand for high-precision alignment during the hoisting process of bridge-erecting machines, this paper presents a two-stage marker detection–localization network tailored to hoisting alignment. The proposed network adopts a “coarse detection–fine estimation” phased framework; the first stage employs a lightweight detection module, which integrates a dynamic hybrid backbone (DHB) and dynamic switching mechanism to efficiently filter background noise and generate coarse localization boxes of marker regions. Specifically, the DHB dynamically switches between convolutional and Transformer branches to handle features of varying complexity (using depthwise separable convolutions from MobileNetV3 for low-level geometric features and lightweight Transformer blocks for high-level semantic features). The second stage constructs a Transformer-based homography estimation module, which leverages multi-head self-attention to capture long-range dependencies between marker keypoints and the scene context. By integrating enhanced multi-scale feature interaction and position encoding (combining the absolute position and marker geometric priors), this module achieves the end-to-end learning of precise homography matrices between markers and hoisting equipment from the coarse localization boxes. To address data scarcity in construction scenes, a multi-dimensional data augmentation strategy is developed, including random homography transformation (simulating viewpoint changes), photometric augmentation (adjusting brightness, saturation, and contrast), and background blending with bounding box extraction. Experiments on a real bridge-erecting machine dataset demonstrate that the network achieves detection accuracy (mAP) of 97.8%, a homography estimation reprojection error of less than 1.2 mm, and a processing frame rate of 32 FPS. Compared with traditional single-stage CNN-based methods, it significantly improves the alignment precision and robustness in complex environments, offering reliable technical support for the precise control of automated hoisting in bridge-erecting machines. Full article

Remote Sighted Assistance (RSA) systems provide visually impaired people (VIPs) with real-time guidance by connecting them with remote sighted agents to facilitate daily travel. However, unfamiliar environments often complicate decision-making for agents and can induce anxiety in VIPs, thereby reducing the effectiveness of the assistance provided. To address this challenge, this paper proposes a video-based map assistance method. By pre-recording pedestrian path videos and aligning them with geographic locations, the system enables route preview and enhances navigation guidance. This study introduces a factor graph optimization (FGO) algorithm that integrates Global Navigation Satellite System (GNSS) and pedestrian dead reckoning (PDR) data for pedestrian positioning. It incorporates road-anchor constraints, a turning-point-based anchor-matching method, and a coarse-to-fine optimization strategy to improve the positioning accuracy. GNSS provides global reference positions, PDR offers precise relative motion constraints through accurate heading estimation, and anchor factors further enhance localization accuracy by leveraging known geometric features. We collected data using a smartphone equipped with a four-camera module and conducted tests in representative urban environments. Experimental results demonstrate that the proposed anchor-aided FGO-GNSS/PDR algorithm achieves robust and accurate positioning, effectively supporting video-based map construction in complex urban settings. With anchor constraints, the mean horizontal positioning error was reduced by 42% to 65% and the maximum error by 38% to 76% across all datasets. In this study, the mean horizontal positioning error was 1.36 m. Full article

High-performance concrete (HPC) is an essential construction material used for modern buildings and infrastructure assets, recognized for its exceptional strength, durability, and performance under harsh situations. Nonetheless, the HPC production process frequently correlates with elevated carbon emissions, principally attributable to the high quantity of cement utilized, which significantly influences its carbon footprint. In this study, data-driven modeling and optimization strategies are employed to minimize the carbon footprint of high-performance concretes while keeping their performance properties. Starting from an experimental dataset, artificial neural networks (ANNs), ensemble techniques (ETs), and Gaussian process regression (GPR) are employed to yield predictive models for compressive strength of HPC mixes. The model’s input variables are the various components of HPC: cement, water, superplasticizer, fly ash, blast furnace slag, and coarse and fine aggregates. Models are trained using a dataset of 356 records. Results proved that the GPR-based model exhibits excellent accuracy with a determination coefficient of 0.90. The prediction model is used in a double objective optimization task formulated to identify mix configurations that allow for high mechanical performance aligned with a reduced carbon emission. The multi-objective optimization task is undertaken using genetic algorithms (GAs). Promising results are obtained when the machine learning prediction model is associated with GA optimization to identify strong yet sustainable mix configurations. Full article

Plant disease segmentation has achieved significant progress with the help of artificial intelligence. However, deploying high-accuracy segmentation models in resource-limited settings faces three key challenges, as follows: (A) Traditional dense attention mechanisms incur quadratic computational complexity growth (O($n^{2}d$)), rendering them ill-suited for low-power hardware. (B) Naturally sparse spatial distributions and large-scale variations in the lesions on leaves necessitate models that concurrently capture long-range dependencies and local details. (C) Complex backgrounds and variable lighting in field images often induce segmentation errors. To address these challenges, we propose Sparse-MoE-SAM, an efficient framework based on an enhanced Segment Anything Model (SAM). This deep learning framework integrates sparse attention mechanisms with a two-stage mixture of experts (MoE) decoder. The sparse attention dynamically activates key channels aligned with lesion sparsity patterns, reducing self-attention complexity while preserving long-range context. Stage 1 of the MoE decoder performs coarse-grained boundary localization; Stage 2 achieves fine-grained segmentation by leveraging specialized experts within the MoE, significantly enhancing edge discrimination accuracy. The expert repository—comprising standard convolutions, dilated convolutions, and depthwise separable convolutions—dynamically routes features through optimized processing paths based on input texture and lesion morphology. This enables robust segmentation across diverse leaf textures and plant developmental stages. Further, we design a sparse attention-enhanced Atrous Spatial Pyramid Pooling (ASPP) module to capture multi-scale contexts for both extensive lesions and small spots. Evaluations on three heterogeneous datasets (PlantVillage Extended, CVPPP, and our self-collected field images) show that Sparse-MoE-SAM achieves a mean Intersection-over-Union (mIoU) of 94.2%—surpassing standard SAM by 2.5 percentage points—while reducing computational costs by 23.7% compared to the original SAM baseline. The model also demonstrates balanced performance across disease classes and enhanced hardware compatibility. Our work validates that integrating sparse attention with MoE mechanisms sustains accuracy while drastically lowering computational demands, enabling the scalable deployment of plant disease segmentation models on mobile and edge devices. Full article

This study investigates the mechanical performance of concrete paving stones manufactured with recycled aggregates derived from TransMilenio slab demolition waste (CDW-A-TS) as a sustainable alternative to conventional natural coarse aggregates (river gravel) and fine aggregates (river sand). Construction and demolition waste from Bogotá’s mass transit system slabs was processed to produce recycled aggregates, which were replaced at substitution levels of 0%, 30%, 50%, and 100% by volume of natural aggregates. The mechanical properties evaluated included compressive strength, flexural strength, abrasion resistance, and water absorption, following Colombian Technical Standards (NTC) and international protocols. Results demonstrate that all CDW-A-TS mixtures exhibit enhanced compressive strength, with improvements ranging from 14.71% to 32.82% compared to the control mix. Flexural strength also increased by 1.34% to 6.13%. However, water absorption increased proportionally with CDW-A-TS content (10.66% to 25.24%). The optimal substitution level was identified at 30% CDW-A-TS based on a composite evaluation of mechanical performance (compressive and flexural strength), durability indicators (water absorption and abrasion resistance), This research demonstrates the technical viability of incorporating TransMilenio demolition waste in paving stone production, contributing to circular economy principles and sustainable urban infrastructure development. This finding aligns with prior research affirming the viability of incorporating recycled coarse aggregates in concrete prefabricates, such as paving stones, for various construction applications. Full article

With the continuous increase in mining activities, effective tailings management has become a critical concern in geotechnical and environmental engineering. This study systematically investigates the microstructural characteristics and 3D reconstruction behavior of copper tailings with different particle sizes using X-ray computed tomography (micro-CT), digital image processing, and 3D modeling techniques. Two particle size groups (fine: 0.075–0.15 mm; coarse: 0.15–0.3 mm) were analyzed to quantify differences in particle morphology, pore structure, and orientation anisotropy. Binary images and reconstructed models revealed that coarse particles tend to have more irregular and angular shapes, while fine particles exhibit more complex pore networks with higher fractal dimensions. The apparent porosity derived from CT data was consistently lower than laboratory measurements, likely due to internal agglomeration effects. Orientation analysis indicated that particle alignment and anisotropy vary systematically with section angle relative to the principal stress direction. These findings offer new insights into the particle-scale mechanisms affecting the packing, porosity, and anisotropy of tailings, providing a scientific basis for enhancing the structural evaluation and sustainable management of tailings storage facilities. Full article

Accurate registration of multi-platform light detection and ranging (LiDAR) point clouds is essential for detailed forest structure analysis and ecological monitoring. In this study, we developed a novel two-stage method for aligning terrestrial and unmanned aerial vehicle LiDAR point clouds in forest environments. The method first performs coarse alignment using canopy-level digital surface models and Fast Point Feature Histograms, followed by fine registration with Iterative Closest Point. Experiments conducted in six forest plots achieved an average registration accuracy of 0.24 m within 5.14 s, comparable to manual registration but with substantially reduced processing time and human intervention. In contrast to existing tree-based methods, the proposed approach eliminates the need for individual tree segmentation and ground filtering, streamlining preprocessing and improving scalability for large-scale forest monitoring. The proposed method facilitates a range of forest applications, including structure modeling, ecological parameter retrieval, and long-term change detection across diverse forest types and platforms. Full article

The primary challenge that robots face in disaster rescue is to precisely and efficiently construct disaster maps and achieve autonomous navigation. This paper proposes a method for air–ground collaborative map construction. It utilizes the flight capability of an unmanned aerial vehicle (UAV) to achieve rapid three-dimensional space coverage and complex terrain crossing for rapid and efficient map construction. Meanwhile, it utilizes the stable operation capability of an unmanned ground vehicle (UGV) and the ground detail survey capability to achieve precise map construction. The maps constructed by the two are accurately integrated to obtain precise disaster environment maps. Among them, the map construction and positioning technology is based on the FAST LiDAR–inertial odometry 2 (FAST-LIO2) framework, enabling the robot to achieve precise positioning even in complex environments, thereby obtaining more accurate point cloud maps. Before conducting map fusion, the point cloud is preprocessed first to reduce the density of the point cloud and also minimize the interference of noise and outliers. Subsequently, the coarse and fine registrations of the point clouds are carried out in sequence. The coarse registration is used to reduce the initial pose difference of the two point clouds, which is conducive to the subsequent rapid and efficient fine registration. The coarse registration uses the improved sample consensus initial alignment (SAC-IA) algorithm, which significantly reduces the registration time compared with the traditional SAC-IA algorithm. The precise registration uses the voxelized generalized iterative closest point (VGICP) algorithm. It has a faster registration speed compared with the generalized iterative closest point (GICP) algorithm while ensuring accuracy. In reinforcement learning navigation, we adopted the deep deterministic policy gradient (DDPG) path planning algorithm. Compared with the deep Q-network (DQN) algorithm and the A* algorithm, the DDPG algorithm is more conducive to the robot choosing a better route in a complex and unknown environment, and at the same time, the motion trajectory is smoother. This paper adopts Gazebo simulation. Compared with physical robot operation, it provides a safe, controllable, and cost-effective environment, supports efficient large-scale experiments and algorithm debugging, and also supports flexible sensor simulation and automated verification, thereby optimizing the overall testing process. Full article

Lane detection plays a fundamental role in autonomous driving systems, yet it remains challenging under complex real-world conditions such as low illumination, occlusion, and degraded lane markings. In this paper, we propose a novel lane detection framework, Informative Feature Pyramid Network (Info-FPNet), designed to improve multi-scale feature representation and alignment for robust lane detection. Specifically, the proposed architecture integrates two key modules: an informative feature pyramid (IFP) module and a cross-layer refinement (CLR) module. The IFP module selectively aggregates spatially and semantically informative features across different scales using pixel shuffle upsampling, feature alignment, and semantic encoding mechanisms, thereby preserving fine-grained details and minimizing aliasing effects. The CLR module applies region-wise attention and anchor regression to refine coarse lane proposals, enabling better localization of curved or occluded lanes. Experimental results on two public benchmarks, CULane and TuSimple, demonstrate that the proposed Info-FPNet outperforms state-of-the-art approaches in terms of F1 score and is robust under challenging conditions such as nighttime, strong reflections, and occlusions. Furthermore, the proposed method maintains real-time inference speed and low computational overhead, validating its effectiveness and practicality in real-world applications. Full article