Search Results (47)

Elkhorn coral (Acropora palmata) is a threatened reef-building species that plays a critical role in Caribbean coastal ecosystems. Efficient, large-scale monitoring of A. palmata is essential for evaluating restoration success, yet traditional in situ surveys remain costly and spatially constrained. In this study, we acquired high-resolution (1.8 cm) uncrewed aerial vehicle (UAV) imagery of a coral reef within the United States Virgin Islands’ (USVI) St. Croix East End Marine Park (STXEEMP) and applied deep learning object detection to identify individual A. palmata colonies. We utilized two convolutional neural network architectures, FasterRCNN and MaskRCNN. FasterRCNN was used as an initial screening tool to identify the optimal imagery dataset from several candidates. After identifying the dataset, we used MaskRCNN with an iterative annotation refinement procedure in which initial model predictions were used to augment the training data and achieved an F1 score of 0.78. Detection accuracy was strongly influenced by colony size and apparent water depth, with markedly high accuracy for corals wider than 0.3 m (F1 = 0.87) and located in shallower waters (F1 = 0.81). Beyond detection, MaskRCNN’s polygon outputs enabled the measurement of the individual colony area and the generation of high-resolution coral density maps. These products complement broader-scale prediction and mapping approaches and provide fine-scale, management-relevant information. Although this study was conducted at a single reef site during one acquisition period, the results suggest that UAV-based deep learning workflows offer a promising approach for coral reef monitoring that could support restoration assessments and conservation decision-making, pending validation across additional sites, seasons, and environmental conditions. Full article

While infrastructure asset management increasingly relies on high-resolution drone imagery, existing workflows suffer from fragmented information management and dependence on costly local processing infrastructure. This paper addresses these limitations by using a cloud-native spatial intelligence hub that converts raw inspection imagery into an interactive and queryable three-dimensional information layer. The system integrates a timeout-resilient orchestration layer for photogrammetry pipelines, a multi-user three-dimensional environment for collaborative review, and a PostGIS-backed spatial database that stores defects as georeferenced anchors. We further introduce a spatial anchoring workflow mapping three-dimensional interactions to world coordinates, retrieving context-relevant images via frustum-based visibility scoring. Evaluated on real inspection datasets, the serverless architecture achieved end-to-end reconstruction in under one hour with sub-25 ms query latency. Results indicate that acquisition geometry, particularly oblique convergent viewpoints, is a stronger predictor of reconstruction complexity than image count. This work establishes a reproducible reference architecture, enabling a transition from file-centric documentation to traceable, spatially indexed evidence management for infrastructure Digital Twins. Full article

Deep learning-based automatic target recognition (ATR) for synthetic aperture radar (SAR) imagery requires large quantities of high-quality annotated data, yet real SAR samples are costly and difficult to obtain. Existing generative adversarial network (GAN)-based SAR generation methods often suffer from limited authenticity and insufficient diversity. To address these issues, we propose CA-DDPM, a conditionally embedded attention-aided denoising diffusion probabilistic model (DDPM) for high-quality multi-category SAR image generation. CA-DDPM employs a unified conditional embedding that fuses time-step and category information, injected into a U-Net backbone through a feature-wise linear modulation (FiLM)-based mechanism to achieve step-aware and class-aware denoising. Attention blocks are further incorporated to enhance the modeling of structural dependencies and fine scattering details. To evaluate generation quality, we develop a three-dimensional assessment framework that jointly examines authenticity, diversity, and utility in ATR. Authenticity is quantified using local and global similarity metrics under a unified Hungarian-matched statistical procedure, together with an SAR-adapted Fréchet inception distance (SAR-FID). Diversity is assessed through inter-category feature clustering, an SAR Inception Score (SAR-IS), and a newly proposed intra-category grayscale histogram-based metric. Utility is evaluated by hybrid training experiments across multiple ATR models. Experiments on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset demonstrate that CA-DDPM produces more realistic and diverse SAR images than representative GAN- and DDPM-based baselines, and it effectively improves downstream ATR performance through data augmentation. Full article

Synthetic aperture radar (SAR) is of great value in intelligent image interpretation. However, the acquisition of real SAR data is costly, and manual annotation heavily relies on expert experience. These factors severely restrict the development of SAR intelligent interpretation algorithms. Meanwhile, the high-frequency details of SAR images contain rich target information. Traditional generation methods cannot effectively capture these key features. To address the above issues, this paper proposes a dual-branch feature decoupling generative adversarial network (GAN) with wavelet constraint designed to achieve high-quality and parameter-controllable SAR image generation. The framework leverages discrete wavelet transform (DWT) to separate spatial structure from high-frequency details, which are independently modeled by a structure branch and a detail branch, respectively. A wavelet consistency loss function is introduced to constrain the distribution of generated and real images in high-frequency subbands, thereby enhancing the model’s capability to model scattering details. To fuse features from the two branches, a cross-attention fusion module is adopted to realize the adaptive compensation of structural features with texture details. Furthermore, to achieve joint control over the semantic attributes and azimuth of generated samples, the framework further integrates auxiliary classification and azimuth regression tasks. A multi-task learning mechanism is constructed to realize precise control over the target’s semantic category and azimuth. For the continuous variable of azimuth, an angle-aware hypernetwork transform module is introduced to perform dynamic convolution modulation on the structure branch at the feature map scale, which improves the model’s fine control capability over continuous azimuth variations. Experimental results on the MSTAR dataset demonstrate that the proposed model can significantly improve the semantic consistency and visual fidelity of the generated samples. The generated samples exhibit high statistical alignment with real data distributions, confirming the model’s effectiveness in characterizing the feature space of SAR imagery and enabling controllable SAR data simulation, thereby augmenting datasets for image interpretation tasks. Full article

Real-world autonomous agents learn under nonstationarity, safety constraints, and finite energetic budgets. We develop a framework for perennial learning—agents that continuously refine their models while provably controlling the cost of forgetting—by unifying three classical pillars: Kolmogorov complexity, which equates scientific discovery with algorithmic compression; Landauer’s principle, which assigns a minimal thermodynamic cost of $k_{\mathrm{B}}Tln2$ per erased bit to every irreversible model update; and port-Hamiltonian (PH) dynamics, whose $(J-R)\nabla H$ decomposition separates zero-cost reversible inference from costly irreversible forgetting by construction. The Maxwell demon analogy is formalized: each learning episode is a Szilard cycle in which information acquisition, belief transport, and memory erasure must balance thermodynamically. The information-distance framework, comprising the normalized information distance (NID) and normalized compression distance (NCD), provides a computable geometry for measuring learning progress and guiding curriculum design. We separate theideal uncomputable regularizer based on prefix complexity from the practical compressor/MDL (minimum description length) surrogate that appears in optimization and prove a calibration lemma linking the two under a mild uniform-accuracy assumption. Under explicit regularity, compact-sublevel, and non-energy-extracting assumptions, we prove a passivity speed limit for curriculum-induced contractions of the effective feasible set. Under local asymptotic normality, we reprove that Fisher information is a local posterior codelength proxy rather than an exact theorem about algorithmic entropy. A conditional sequential information-budget proposition shows that the per-stage sample requirement scales as $\tilde{O}(\Delta k_t/{\lambda }_{\star })$, where $\Delta k_t$ is the number of materially changed model coordinates (not the total model complexity $k_t$); the $k^3\to \Delta k$ improvement is conditional on a warm-start assumption and a chosen cold-start baseline. A double-integrator running example with a moving obstacle illustrates the architecture. Full article

The concentrations of nitrogen (N), phosphorus (P), potassium (K), organic matter (OM), and pH in soil are critical markers of fertility that influence crop growth and yield. Traditional wet-chemical analyses are labor-intensive, time-consuming, and costly, thereby constraining timely soil information acquisition for precision agriculture. This study evaluates whether multi-output Gaussian process regression (MOGPR) can enhance the prediction accuracy of multiple soil nutrients by exploiting their intrinsic correlations, in comparison with single-output Gaussian process regression (SOGPR). Near-infrared (NIR) spectroscopy was applied to 622 typical black soil samples collected from the Farm 855 (45°43′ N, 131°35′ E), Heilongjiang Province, China. Corresponding MOGPR and SOGPR models were developed for systematic performance comparison. Results indicated that MOGPR significantly outperformed SOGPR for nutrients exhibiting moderate-to-strong intercorrelations (N, P, K, and OM), yielding R² improvements of 0.070.28 and RPD increases of 16–40%, whereas only limited gains were observed for pH due to its weak correlations with other nutrients. These findings indicate that combining NIR spectroscopy with MOGPR offers significant potential for rapid, nondestructive assessment of multiple soil nutrients. This study further establishes a correlation-aware multi-output modeling framework that links shared spectral responses with an inter-nutrient dependency structure, providing methodological guidance for multi-nutrient soil prediction. Full article

Soil texture is recognised as one of the key physical properties of soil. Although traditional laboratory testing methods can determine soil texture information with high accuracy, they are often considered time-consuming and costly. To achieve rapid and accurate acquisition of soil texture information, this study proposes RVFM, a hybrid deep learning model designed for soil texture detection using microscopic images. The model integrates a CNN branch for extracting multi-dimensional texture features with a Transformer branch for capturing global positional information, fused via a cross-attention module. This architecture effectively captures microscopic distribution characteristics to estimate soil composition proportions. Experimental results demonstrate high precision, with prediction coefficients (R²) for sand, silt, and clay reaching 0.971, 0.954, and 0.931, respectively. Corresponding Root Mean Square Errors (RMSE) were recorded at 3.789, 2.842, and 2.780. The test results outperform those of other classical network models, and the model shows better fitting performance in generalisation tests, demonstrating certain practical value Full article

Reliable assessments of dam stability require the continuous acquisition and interpretation of deformation data, as monitoring technologies provide essential information for evaluating structural behavior. Surface displacement measurements are particularly valuable for identifying instability within the dam embankment and adjacent slopes. While terrestrial surveying networks can provide accurate point-based observations, they are often time-consuming and costly to maintain. Satellite radar interferometry (InSAR) offers a complementary, cost-effective means of monitoring surface displacement with wide spatial coverage; however, careful analysis is required to avoid misinterpreting superficial motions of riprap and cover materials as true dam settlement. In this study, we use multi-platform SAR datasets, including Sentinel-1A (2014–2019) and high-resolution TerraSAR-X (2018), to investigate the deformation behavior of the Taleqan Dam. We compare LOS displacement derived from InSAR with independent measurements from a terrestrial surveying network spanning the same period. TerraSAR-X data indicate up to ~20 mm of LOS displacement over three months (May–August 2018), and the displacement pattern is consistent with the Sentinel-1 time series. Despite lower spatial resolutions, Sentinel-1 provided dense, temporally continuous coverage, with LOS velocities reaching ~4 mm/yr on the downstream slope. The combined datasets demonstrate that the observed deformation predominantly reflects the ongoing lateral movement of downstream riprap materials rather than the vertical settlement of the dam’s core. These results highlight both the utility of InSAR for long-term dam monitoring and the importance of integrating multi-sensor observations to ensure accurate interpretations of dam deformation signals. Full article

We introduce Causal Incentive Design (CID), a framework that applies causal inference to canonical single-stage principal–agent problems (PAPs) characterized by bilateral private information. Within CID, the operating rules of PAPs are formalized using an additive-noise causal graphical model (CGM). Incentives are modeled as interventions on a function space variable, $\Gamma$, which correspond to policy interventions in the principal–follower causal relation. The causal inference target estimand $V(\Gamma )$ is defined as the expected value of the principal’s utility variable under a specified policy intervention in the post-intervention distribution. In the context of additive-Gaussian independent noise, the estimand $V(\Gamma )$ decomposes into a two-layer expectation: (i) an inner Gaussian smoothing of the principal’s utility regression; and (ii) an outer averaging over the conditional probability of the follower’s action given the incentive policy. A Gauss–Hermite quadrature method is employed to efficiently estimate the first layer, while a policy-local kernel reweighting approach is used for the second. For offline selection of a single incentive policy, a Functional Causal Bayesian Optimization (FCBO) algorithm is introduced. This algorithm models the objective functional $\gamma \mapsto V\left(\gamma \right)$ using a functional Gaussian process surrogate defined on a Reproducing Kernel Hilbert Space (RKHS) domain and utilizes an Upper Confidence Bound (UCB) acquisition functional. Consequently, the policy value $V\left(\gamma \right)$ becomes an interventional query that can be answered using offline observational data under standard identifiability assumptions. High-probability cumulative-regret bounds are established in terms of differential information gain for the proposed FBO algorithm. Collectively, these elements constitute the central contributions of the CID framework, which integrates causal inference through identification and estimation with policy search in principal–agent problems under private information. This approach establishes a causal decision-making pipeline that enables commitment to a high-performing incentive in a single-shot game, supported by regret guarantees. Provided that the data used for estimation is sufficient, the resulting offline pipeline is appropriate for scenarios where adaptive deployment is impractical or costly. Beyond the methodological contribution, this work introduces a novel application of causal graphical models and causal reasoning to incentive design and principal–agent problems, which are central to economics and multi-agent systems. Full article

With the continuous development of cities and the increasing utilization of underground space, ambient noise seismic imaging has become an essential approach for exploring and monitoring the urban subsurface. The integration of Distributed Acoustic Sensing (DAS) with ambient noise imaging offers a more convenient and effective solution for investigating shallow subsurface structures in urban environments. To overcome the limitations of conventional ambient noise seismic nodes, which are costly and incapable of achieving high-density data acquisition, this work makes use of existing urban telecommunication fibers to record ambient noise and perform sliding-window cross-correlation on it. Then the Phase-Weighted Stack (PWS) technique is applied to enhance the quality and stability of the cross-correlation signals, and fundamental-mode Rayleigh wave dispersion curves are extracted from the cross-correlation functions through the High-Resolution Linear Radon Transform (HRLRT). In the inversion stage, an adaptive regularization strategy based on automatic L-curve corner detection is introduced, which, in combination with the Preconditioned Steepest Descent (PSD) method, enables efficient and automated dispersion inversion, resulting in a well-resolved near-surface S-wave velocity structure. The results indicate that the proposed workflow can extract useful surface-wave dispersion information under typical urban noise conditions, achieving a feasible level of subsurface velocity imaging and providing a practical technical means for urban underground space exploration and utilization. Full article

Machine learning approaches are commonly used to model physical phenomena due to their adaptability to complex systems. In general, a substantial number of samples must be collected to create a model with reliable results. However, collecting numerous data points is often costly. Moreover, high-dimensional problems inherently require large amounts of data due to the curse of dimensionality. That is why new approaches based on smart sampling techniques are being investigated to optimize the acquisition of training samples, such as active learning methods. Initialization is a crucial step in active learning as it influences both performance and computational cost. Moreover, the scenarios used to select the next sample, such as classic pool-based sampling, can be highly resource- and time consuming. This study focuses on optimizing active learning methods through a comprehensive analysis of initialization strategies and scenario design, proposing and evaluating multiple approaches to determine the optimal configurations. The methods are applied to high-dimensional industrial problems with dimensions ranging from 5 to 15, where challenges associated with high dimensionality are already significant. To address this, the proposed study uses an active learning criterion that combines Sparse Proper Generalized Decomposition with Fisher information theory, specifically tailored to high-dimensional industrial settings. We illustrate the effectiveness of these techniques through examples on theoretical 5D and 15D functions, as well as a practical industrial crash simulation application. Full article

Hyperspectral images (HSIs) provide rich spectral–spatial information and support applications in remote sensing, agriculture, and medicine, yet their development is hindered by data scarcity and costly acquisition. Diffusion models have enabled synthetic HSI generation, but conventional integer-order solvers such as Denoising Diffusion Implicit Models (DDIM) and Pseudo Linear Multi-Step method (PLMS) require many steps and rely mainly on local information, causing error accumulation, spectral distortion, and inefficiency. To address these challenges, we propose Integer–Fractional Alternating Diffusion Sampling (IFADiff), a training-free inference-stage enhancement method based on an integer–fractional alternating time-stepping strategy. IFADiff combines integer-order prediction, which provides stable progression, with fractional-order correction that incorporates historical states through decaying weights to capture long-range dependencies and enhance spatial detail. This design suppresses noise accumulation, reduces spectral drift, and preserves texture fidelity. Experiments on hyperspectral synthesis datasets show that IFADiff consistently improves both reference-based and no-reference metrics across solvers without retraining. Ablation studies further demonstrate that the fractional order $\alpha$ acts as a controllable parameter: larger values enhance fine-grained textures, whereas smaller values yield smoother results. Overall, IFADiff provides an efficient, generalizable, and controllable framework for high-quality HSI generation, with strong potential for large-scale and real-time applications. Full article

The reliability of electromechanical systems is a critical factor in modern Industry 4.0, as unexpected failures in induction motors or gearboxes can cause costly downtime, productivity losses, and increased maintenance demands. Infrared thermography offers a non-invasive and real-time means of monitoring thermal behavior, yet its effective use for fault diagnosis remains challenging due to sensitivity to noise, environmental variability, and the need for robust feature extraction. This work proposes a novel end-to-end convolutional neural network (CNN) methodology for detecting and classifying faults in electromechanical systems through the processing of infrared thermography images. The method integrates an automatic preprocessing stage that isolates the Relevant Heating Areas (RHAs), preserving their geometric and thermal descriptors while filtering irrelevant background information. A tailored data augmentation strategy, including controlled noise injection, was designed to improve robustness under realistic acquisition conditions. The CNN architecture combines 3 × 3 and 5 × 5 kernels to capture both fine-grained and global heating patterns. Experimental validation is carried out under nine different faulty conditions, achieving 99.7% accuracy and demonstrating strong resilience against Gaussian blur and additive Gaussian noise. The results suggest that the method provides a scalable, interpretable, and efficient approach for fault diagnosis in electromechanical systems within Industry 4.0 environments. Full article

To build robust condition monitoring solutions, it is important to identify signals that capture relevant information. However, how a degradation affects a given part of machinery might not be clear at the beginning. As a result, exploration measurement campaigns collecting large amounts of data are needed for initial evaluation. Vibration signals are typical examples of such data. Although, for explorative measurement campaigns, the battery-powered wireless node brings extra flexibility in terms of positioning the sensor at the desired location and facilitates retrofitting, the limited energy posed by them is the major downside. Sending high-sampled data over wireless channels is costly energy-wise if all samples are to be sent. When multiple sensor nodes transmit real-time measurement data concurrently over a wireless channel, the risk of channel saturation increases significantly. Avoiding this requires identifying an optimal balance between sampling time, transmission duration, and payload size. This can be done by processing and compressing data before transmission, on the sensor node close to the data acquisition and later reconstructing the received samples on the central node. In this paper, we analyze two compression mechanisms to ensure a good compression ratio and still allow good signal reconstruction for later analysis. We study two approaches, one based on the Fast Fourier Transform and one on Singular Value Decomposition, and discuss the pros and cons of each variant. Full article

Monitoring the condition of existing structures remains one of the most pressing challenges within the construction industry. Structural health monitoring (SHM) techniques have proven increasingly effective in this regard; however, maintaining and archiving complete lifecycle data for such structures remains costly. Data acquisition is particularly critical, as the SHM system relies upon this information to analyse and evaluate structural behaviour. Nonetheless, a range of challenges—such as environmental influences, sensor malfunction, and transmission failures—can lead to data corruption or loss. These issues compromise the reliability of the dataset, necessitating either data reconstruction or additional measurement campaigns, both of which are resource-intensive. This study proposes the use of a long short-term memory (LSTM) network to reconstruct missing or corrupted data. A complete dataset collected from an actual construction project is employed to train the network. Data loss scenarios are then simulated, including single-channel (loss from one sensor) and multi-channel (loss from multiple sensors) cases. The trained LSTM model is subsequently applied to reconstruct the missing data. A case study on a real bridge demonstrates that the reconstructed data show strong agreement with the original measurements in both the time and frequency domains. These findings indicate that the proposed approach has the potential to support engineers in conserving resources by reducing the need for costly and time-consuming additional measurement interventions. Full article