Search Results (11,962)

Diffractive optical elements (DOEs) provide an ultra-lightweight route to large-aperture imaging and are therefore attractive for lightweight large-aperture remote sensing. However, images directly captured by DOE-based diffraction systems often suffer from blur and low contrast caused by order aliasing, spatially varying blur, and uncertain noise levels. These degradations reduce image interpretability in remote sensing scenes, while the lack of large-scale paired diffraction datasets further limits data-driven restoration. To address this problem, we propose PLDIR, a diffraction remote sensing image restoration framework that integrates maximum a posteriori (MAP) estimation with a latent diffusion prior. PLDIR reduces manual noise-level tuning and alleviates the limitation of globally uniform denoising strength in conventional plug-and-play restoration. The framework contains three stages. Stage 1 trains a latent relationship encoder (LRE) to capture the latent relationship $H_{gt}$ between clean and noisy images. Stage 2 freezes the trained LRE and learns a latent diffusion denoising model (LDDM) to predict $H_{prev}$ from noisy images, providing a noise-aware prior for adaptive denoising. Stage 3 embeds the LDDM into the half-quadratic splitting (HQS) framework, where $H_{prev}$ regulates both the denoising strength and the spatially adaptive penalty matrix during iteration. Experiments on synthetic diffraction remote sensing data and real captured diffraction images demonstrate that PLDIR improves restoration quality and detail preservation over representative baselines, providing an effective restoration approach for lightweight diffraction remote sensing. Full article

Despite the importance of adaptive reuse for urban renewal, municipal records of changes in property destinations remain an untapped resource for understanding regulatory barriers. This study develops an interpretable machine learning (ML) framework to screen historical administrative non-approval using 42,350 municipal records from a mid-sized city in southern Chile spanning 2010 to 2025. Historical memory features were constructed exclusively from outcomes available before each application, thereby preventing temporal leakage. Records from 2010–2022 were used for training, 2023 was reserved for validation, and 2024–2025 constituted a locked test set. A logistic regression model combining auditable risk flags derived from GIRO (declared economic or functional activity) with historical property, address, and activity category information was selected by maximizing the F1-score (the harmonic mean of precision and recall) subject to a recall ≥ 0.85. At the validation-selected threshold of 0.285, the model achieved a receiver operating characteristic area under the curve (ROC-AUC) of 0.778, a precision–recall area under the curve (PR-AUC) of 0.720, an F1-score of 0.719, recall of 0.887, and precision of 0.604. The model flagged 66.8% of applications for prioritized review. The proposed framework supports human-in-the-loop screening and institutional learning; it neither determines regulatory compliance nor automates municipal decisions. Full article

Electrode-level compressibility may not imply transferable hardware redundancy in biomimetic myoelectric interfaces. This study tested whether sensor-count sufficiency discovered by trial-level analysis survives participant-grouped evaluation. Dataset A comprised 398 archived Myo Armband trials from eight gestures. Dataset B contained 864 one-second trials from 36 participants and six gestures. A timestamp audit identified extensive repeated channel values; Dataset B was therefore analyzed on a conservative 100 Hz grid with 20–45 Hz filtering. Sensor subsets and RBF-SVM parameters were selected exclusively within grouped training data using repeated nested validation. Electrode-level NMF, all 28 fixed six-sensor layouts, cyclic re-indexing, channel-block ablation, participant-cluster bootstrap, PCA, and time-domain-only sensitivity analyses were evaluated. Dataset A yielded 97.74% accuracy with six sensors and 97.93% with eight. In Dataset B, accuracy was 75.96% ± 7.47% with six sensors and 77.93% ± 6.49% with eight; the paired difference was −1.97 percentage points (corrected 95% CI, −5.50 to 1.56). The participant-cluster bootstrap interval was −3.70 to −0.31 points. Active-gesture accuracy was 71.67% and 74.35%, respectively. All fixed six-sensor layouts averaged 74.59%. Three NMF components reconstructed 89.95% ± 1.78% of held-out-participant normalized RMS patterns, with no nonconverged folds. One-position cyclic re-indexing reduced accuracy to 40.28%; channel-block ablation caused losses of 0.62–6.71 points. Low-dimensional electrode-level RMS structure did not establish removable sensors across unseen users. Compact biomimetic interfaces require registration, adaptation, or equivariant processing before physical sensor reduction. Full article

Hyperspectral images (HSIs) are inevitably degraded by mixed noise, which hampers downstream interpretation. Recent graph-signal-processing denoisers encode the spatial–spectral structure of an HSI through a product graph over superpixel bodies, yet the adopted Kronecker (tensor) product graph retains only the joint spatial–spectral edges and discards the pure-spatial and pure-spectral edges—the two priors that govern HSI smoothness. We introduce a two-parameter weighted product-graph family that contains the Kronecker, Cartesian and strong products as exact special cases, and propose Weighted Strong Product Graph Laplacian Regularization (WSPGLR)—the strong-product branch with a tunable joint-edge weight β—embedded in a global low-rank plus sparse model solved by the Alternating Direction Method of Multipliers (ADMM) with singular-value-thresholding and soft-thresholding updates and a sparse conjugate-gradient (CG) solve. On three simulated cubes (Washington DC Mall, Pavia University, Indian Pines) under four mixed-noise scenarios, WSPGLR consistently improves Mean Peak Signal-to-Noise Ratio (MPSNR) and ERGAS (Erreur Relative Globale Adimensionnelle de Synthèse) over a matched Kronecker-product control in every tested case (and Mean Structural Similarity (MSSIM) in 11 of 12 settings)—up to 2.5 dB in MPSNR from the graph term alone—and an ablation shows that β governs a spatial–spectral fidelity trade-off (Spectral Angle Mapper, SAM). On the detailed urban scenes, WSPGLR attains the highest MPSNR against the external methods and matched control in seven of eight settings, whereas it trails LRTDTV on the smooth agricultural scene; with an optional scene-adaptive TV step it attains the best average rank across scene types; a Tucker-based variant further shows the low-rank block is modular and generally improves spectral fidelity. Tests on four no-reference real HSIs and 30 paired real-noise MEHSI samples extend the sensor coverage; on MEHSI, the native-domain RND framework remains substantially stronger, which delimits the scope of the proposed training-free regularizer. Full article

Seroprevalence surveys constitute a critical tool for estimating population exposure to infectious agents, particularly during public health emergencies. In the context of the coronavirus disease 2019 (COVID-19) pandemic, such surveys revealed that the circulation of Severe Acute Respiratory Syndrome Coronavirus 2 in Africa had been largely underestimated. The validity of these surveys, however, depends not only on the reliability of diagnostic tests but also on operational conditions, which are frequently challenged by fear, misinformation, and logistical constraints. In the Democratic Republic of the Congo, the National Institute of Biomedical Research conducted two seroprevalence surveys in Kinshasa in 2020 on behalf the project “Appui à la Riposte Africaine à l’Epidémie de COVID-19” (ARIACOV). This study described the challenges encountered and the adaptive strategies employed during these surveys, in order to improve future investigations. The study population comprised 41 eligible field workers who participated in the 2020 ARIACOV seroprevalence surveys in Kinshasa, including 28 interviewers and 13 specimen collectors. Of these, 24 individuals (16 interviewers and 8 specimen collectors) initially confirmed their participation. Ultimately, 15 participants took part in the focus group discussions, comprising 11 interviewers (39.3% of the 28 eligible interviewers) and 4 specimen collectors (30.8% of the 13 eligible specimen collectors). We collected data through two focus groups held in a single session, using a pre-tested semi-structured guide. Discussions were audio-recorded and complemented by observational notes until empirical saturation was achieved. We applied deductive thematic analysis, following Braun and Clarke’s framework, to three predefined domains: transportation, survey procedures, and sample management. The application of systematic manual coding, combined with researcher triangulation, reinforced the rigor and enhanced the credibility of the findings. The findings highlighted community mistrust, driven by rumors, apprehension regarding blood collection, financial suspicions, and inadequate communication. In addition, logistical and transportation challenges significantly constrained survey feasibility. Nonetheless, a range of adaptive strategies were implemented, enabling the successful conduct of the surveys despite the pandemic context and limited resources. The participants’ narratives suggest that strengthening field workers’ training, improving transparency, and working alongside community health workers could contribute to enhance the effectiveness of serological surveys during health crises in resource-constrained environments. Full article

Maritime Education and Training (MET) is undergoing rapid digital transformation, yet adoption of adaptive learning technologies (ALT) remains uneven, constrained by the Standards of Training, Certification and Watchkeeping (STCW) regime and by human and organizational barriers. A persistent question in the technology-acceptance literature is how individual acceptance becomes organizational readiness in safety-critical, regulated settings. This study models the translation from educator attitudes to perceived institutional readiness through the mediating role of perceived usefulness. Partial least squares structural equation modeling was applied to survey data from 104 maritime educators. The model explained 42.8% of the variance in perceived institutional readiness and 34.8% in perceived usefulness. Educator attitudes significantly predicted perceived institutional readiness (β = 0.298), and the relationship was partially mediated by perceived usefulness (indirect β = 0.202, 95% CI 0.12 to 0.35): educators must first be convinced of an ALT’s practical value before judging their institution ready to adopt it. A hypothesized moderation by professional experience was not supported and is reported as exploratory. Because predictor and outcome were self-reported by the same respondents, the outcome is interpreted as perceived readiness rather than a verified organizational state. The study contributes a validated model of the individual-to-organizational translation mechanism in MET and guidance for MET leaders integrating ALT under STCW. Full article

Video anomaly detection has two very different answers to the question of how a detector should acquire its notion of “normal” for a given camera: adapt its parameters to hours of normal footage recorded by that exact camera, or perform no target-scene adaptation at all and rely on frozen, heavily pretrained backbones—a vision-language image–text model and a large language model applied to text descriptions of frames. We call the second setting target-scene-training-free: the backbones themselves are trained on very large general-purpose corpora, but no parameter is updated, fine-tuned, or transfer-learned on the target scene. That setting is far more convenient to deploy, but how much accuracy does it give up, and can any of that gap be closed without target-scene adaptation? We define an anomaly operationally, as a frame whose score under a scoring function $s(·)\in [0,1]$ exceeds a threshold, and evaluate every system with one metric definition and one implementation: frame-level area under the ROC curve (AUC) and equal error rate (EER), computed over all ground-truth-labeled test frames of each benchmark. We train a simplified future-frame-prediction network—a U-Net optimized with an intensity and gradient-difference loss, with the optical-flow and adversarial terms of the original design removed—on the normal-only training split of UCSD Ped1, UCSD Ped2, and CUHK Avenue, reaching a mean AUC of 0.848. A target-scene-training-free system starts far behind at 0.567, because its semantic signal is dropped in practice: the vision-language model that would produce it is too slow to run over a full test set. We close this gap in three ways, all evaluated on full test sets: CLIP-Guided Semantic Grounding (CSG), Vision-Language Reasoning (VLR), and Statistically-Calibrated Semantic Grounding (SCSG). Our best method raises mean AUC to 0.653, closing roughly a third of the gap with no target-scene adaptation. The trained model nonetheless leads on every benchmark, and because it is deliberately simplified, the gap we report is a conservative lower bound on the gap a fully engineered trained model would show. We report every result honestly, including a prompt-sensitivity analysis showing how much of the training-free result depends on prompt wording, cases where a more sophisticated method did not beat a simpler one, and a measured explanation of why UCSD Ped1 is the weakest benchmark for the trained model. Full article

As core weighing equipment in the logistics and industrial sectors, the accuracy of truck scales is significantly affected by environmental noise, sensor errors, and nonlinear factors. This paper proposes an adaptive smoothing-constrained broad learning system (PSO-MCC-SCBLS) to enhance the precision and robustness of truck-scale weighing. Particle swarm optimization (PSO) is employed to optimize the number of feature windows, feature nodes, enhancement nodes, and the smoothing coefficient of the SCBLS; the maximum correntropy criterion (MCC) replaces the minimum mean square error (MMSE) criterion for training the output-weight matrix; and a smoothing constraint derived from the physical continuity of the weighing system is introduced to improve generalization in small-sample scenarios. The method was validated on real data from an 8-channel, 40-ton truck scale under a corrected evaluation protocol that uses group-wise five-fold cross-validation, selects all hyperparameters by an inner cross-validation on the training folds only, and matches the effective regularization strength across MCC and non-MCC variants. A complete component ablation over the seven BLS-family variants (BLS, MCC-BLS, SCBLS, PSO-BLS, PSO-SCBLS, PSO-MCC-BLS and the full model) is reported alongside RBLS, CatBoost, CNN_Attention and LSTM, together with anti-interference tests under a composite Gaussian-plus-impulsive disturbance injected in three scenarios: disturbed calibration only (A), disturbed calibration and deployment (B), and disturbed deployment only (C). The results delimit the contribution of each component. Automated structural search is the one component whose benefit is large and consistent, reducing clean-data RMSE by 47.7% over a plain BLS. On clean data the full model does not lead: PSO-MCC-BLS attains the lowest RMSE ($0.0209\times {10}^3$ kg) while the full model records $0.0292\times {10}^3$ kg, indicating that under a leakage-free protocol this calibration task is already close to a low-complexity regime. The advantage of the full model is specific and is reported as such: in Scenario B at a noise ratio of 0.5 it achieves the lowest RMSE ($1.7670\times {10}^3$ kg) and the best Friedman rank among all eleven models ($p<0.05$), whereas at the weaker intensity and in Scenario C the deep-learning baselines lead. Once the effective regularization is matched, the MCC term contributes negligibly on this dataset, so the observed robustness is attributable to the tuned BLS-family model as a whole rather than to correntropy weighting in isolation. Full article

We develop a radial basis function neural network (RBFNN) for dynamic equations on time scales using a Gaussian-type activation generated by the time-scale exponential. The activation ${\varphi }_{\alpha ,c}=e_{\ominus p_{\alpha ,c}}(·,c)$, with $p_{\alpha ,c}\left(\tau \right)=\alpha (\tau -c)$, provides a natural time-scale analogue of the classical Gaussian. We establish two-sided exponential bounds, uniform convergence to the classical Gaussian as the graininess tends to zero, and a localisation–admissibility condition arising from positive regressivity. Training is based on residual minimisation with positive width parametrisation, an analytic Jacobian, and variable projection. Numerical experiments on quantum, clustering, and hybrid time scales, together with two-point boundary value problems and joint parameter identification, demonstrate high accuracy and reliable training. The clearest improvement over the classical Gaussian is observed on hybrid domains. The framework is also applied to irregularly sampled orange-tree growth and theophylline pharmacokinetic data, yielding parameter estimates consistent with recurrence-based and published continuous-model values. The proposed method provides a unified time-scale-adapted global surrogate with admissible-width control and a computable a posteriori error certificate. Full article

With the development of artificial intelligence and automated driving technologies, traffic signal control is evolving toward greater flexibility and faster response. From the perspective of the Transportation Cyber-Physical System (T-CPS), this paper focuses on mixed traffic scenarios involving connected and automated vehicles (CAVs) and human-driven vehicles (HVs). It proposes integrating a Large Language Model (LLM) into signal control: roadside devices perceive traffic states, prompt engineering is constructed, and the LLM is driven to reason and generate control signals. On this basis, a CAV speed guidance algorithm is proposed. Controlled SUMO simulations of a single isolated intersection under ideal V2X communication assumptions show that the proposed method improves delay performance under the tested mixed-traffic conditions. As the CAV penetration rate increases, traffic performance is further improved. Additional experiments under emergency-vehicle priority, road-construction constraints, different traffic-demand levels, perception noise, and different decision intervals and guidance ranges provide simulation-based evidence of training-free scenario adaptability and robustness within the examined scope. Although inference latency and remote-API delays constrain the timely availability of fresh LLM actions, the hard-deadline policy and deterministic fallback mechanism maintain continuous signal execution and favorable traffic performance in the controlled SUMO simulations. Full article

Training under extreme class imbalance (>1:100) remains an open problem in weakly supervised learning. The standard remedy—loss-level reweighting (focal loss, asymmetric loss, class-balanced loss)—is widely adopted, yet its behavior at extreme ratios in Multiple Instance Learning (MIL) is poorly understood. On digital breast tomosynthesis (attention-based pooling over frozen EfficientNet-B3 features), we study the optimization bounds under extreme bag-level MIL imbalance (1:251), intervening at two levels: the loss surface (via reweighting) and the gradient dynamics (via a novel q-calculus gradient modification using the Jackson q-derivative). All three reweighting strategies degrade classification relative to unweighted binary cross-entropy (BCE), monotonically, eliminating the loss surface as the bottleneck. Extended evaluation ($n=20$ seeds) shows q-calculus gradient smoothing matches vanilla BCE ($p=0.632$, Cohen’s $d=0.003$) despite provably reducing gradient variance, establishing an empirical ceiling on the optimization-achievable area under the precision-recall curve (AUPRC) of 0.0912; loss reweighting defines the floor at 0.055. Focal loss is additionally catastrophically miscalibrated (ECE > 0.44 vs. 0.036 for vanilla BCE), a collapse that persists under adaptive binning. In this regime, exceeding the ceiling points to the data-representation level, not the optimizer. We further identify ratio-invariant safety—non-degradation at any imbalance ratio, satisfied by vanilla BCE and q-calculus but violated by all reweighting methods—and give recommendations spanning moderate to extreme imbalance. Full article

Reliable train speed estimation requires wheel-speed and Doppler-radar measurements to be fused with acceleration propagation and sparse trackside positioning. Adaptive maximum correntropy filtering can suppress abnormal speed observations. However, because it weights each innovation relative to the predicted state, an accelerometer error that shifts the prediction may cause valid wheel and radar observations to be rejected. This paper proposes a consistency-aided adaptive maximum correntropy robust filter, termed CA-AMCRF, for asynchronous multi-sensor train speed estimation. The filter uses a five-state bias-augmented model to fuse acceleration, wheel speed, Doppler-radar speed, and sparse balise position measurements. During each direct-speed update, a consistency support mechanism compares the bias-compensated wheel–radar disagreement with their common departure from the predicted speed. The resulting evidence corrects the AMCRF measurement weights, recovering reliable direct-speed information during prediction-side errors while preserving robust downweighting when either direct-speed group is abnormal. Controlled simulations show that the method remains robust to isolated wheel- or radar-speed errors and improves accuracy when an accelerometer error shifts the prediction while both direct-speed groups remain reliable. Physics-based adhesion simulations further show the same benefit when an accelerometer error coexists with adhesion-induced wheel-speed distortion. Full article

Federated learning enables collaborative training across distributed sensor networks (DSNs) without requiring the sharing of raw sensing data. However, local updates can leak private information, and untrusted servers may return incorrect aggregation results. Existing verifiable secure aggregation schemes often incur high costs for resource-constrained devices. To address this, a verifiable and compressed FedAdam aggregation scheme (VCFedAdam) is proposesed. VCFedAdam performs masking, aggregation and verification operations within a low-dimensional sketch space, with the server recovering only the aggregated result for a fixed online set R, which consists of the users that successfully submit valid masked sketches in the current round and is fixed before aggregate recovery. Additionally, a commitment-bound verification (CBV) mechanism is designed to prevent the server from adaptively tampering with the aggregated results. At a compression ratio of $25\%$, experimental results show that VCFedAdam reduces computation overhead by $90.25\%$ and communication overhead by $75.38\%$ compared with traditional secure aggregation. On the CIFAR-10 and MNIST datasets, accuracy drops by only $0.64\%$ and $1.16\%$, respectively. Furthermore, security analysis confirms that VCFedAdam achieves both privacy protection and aggregation verifiability. Full article

Interaction with physical objects plays a key role in the development, assessment, and treatment of cognitive and motor skills across the lifespan. In recent years, this interaction has been enhanced by smart objects, namely everyday items augmented with embedded sensing, processing, and feedback capabilities. These objects enable objective behavioural data collection while preserving natural and engaging human–object interactions. However, the literature remains fragmented and often focused on specific applications or interaction paradigms. This scoping review provides an overview of smart objects developed for clinical and educational contexts, with intended uses including assessment, treatment, training, education, and data collection to support machine learning approaches. Forty-two studies published from 2010 up to the final search date of 29 June 2026 were analysed following the PRISMA-ScR guidelines. The review adopts a design- and hardware-oriented perspective, classifying smart objects according to physical shape, intended use, target population, and validation level. Particular attention is given to embedded electronic components, measured parameters, and the exploitation of sensing and feedback technologies during human–object interaction. By synthesising current solutions, this review highlights emerging trends, recurring limitations, and open challenges related to design choices, technological constraints, and experimental validation, supporting the development of robust, adaptable, and real-world-ready smart objects. Full article

Low-light image super-resolution aims to recover normal-light high-resolution images from dark low-resolution observations captured by image sensors, in which illumination attenuation, sensor noise, blur, and low resolution are entangled, making it more challenging than conventional super-resolution. Diffusion-based methods perform well on real-world super-resolution but usually require costly multi-step inference; recent one-step methods either rely on a globally fixed timestep that cannot adapt to per-sample degradation, or they directly encode the dark image into the latent space, coupling illumination bias with content degradation. To address these issues, we propose RASR, a Retinex-guided adaptive one-step diffusion framework for low-light super-resolution. We first decompose the observation into reflectance and illumination, and we use the reflectance as the content carrier to align it with the normal-light prior of the pretrained model. A latent-space teacher then constructs per-sample supervision from the low/high-quality latent discrepancy, while a lightweight student predicts the noise level solely from the Retinex representation, removing the dependence on high-quality references at inference. Finally, a single velocity-field integration on Stable Diffusion 3 yields the result, updating only low-rank adapters and lightweight modules during training. Extensive experiments on the RELLISUR benchmark show that RASR overall outperforms existing low-light and mainstream super-resolution methods in PSNR, SSIM, and LPIPS, with especially prominent gains in perceptual quality, and ablation studies validate the effectiveness of each key design. Full article