154092 research outputs found
Sort by
Vision on the Move:Automated Hazardous Material Plate Detection in Freight Transport
No full textEnhancing the logistic efficiency and safety of freight transport requires fast, reliable identification of hazardous materials (hazmat). In this work, we explore how computer vision can automate the detection and reading of hazmat number plates on freight trains and trucks. We benchmark two object detection models for hazmat localization, YOLOv11x and Faster R-CNN, across a private freight train dataset and HazTruck, our newly introduced public dataset. For reading the detected plates, we evaluated three Optical Character Recognition (OCR) methods: the widely used Tesseract, EasyOCR, and the recent vision-language model Idefics2. Integrating YOLOv11x and Idefics2 into a unified pipeline achieved the state-of-the-art performance, with over 90% accuracy on freight train data, showcasing a powerful and scalable solution for automated hazmat identification in transport logistics. The code and datasets are available via https://github.com/Robust-Rail.</p
Electric-Field Control of Zero-Dimensional Topological States in Ultranarrow Germanene Nanoribbons
No full textReversible, all-electric control of symmetry-protected zero-dimensional modes has been a long-standing goal. In buckled honeycomb lattices, a perpendicular field couples to the staggered sublattice potential providing the required handle. We combine scanning tunneling microscopy and tight-binding theory to switch zero-dimensional topological end states reversibly on and off in ultranarrow germanene nanoribbons by tuning the electric field in the tunnel junction. Increasing the field switches off the end modes of topological two-hexagon-wide ribbons, while the same field switches on zero-dimensional states in initially trivial three- and four-hexagon-wide ribbons. This atomic scale platform realizes a proof of principle for a zero-dimensional topological field effect device, opening a path for ultrasmall memory, controllable qubits, and neuromorphic architectures.</p
Single-hologram generation of vector beams using a digital micromirror device
No full textWe describe a simple and effective method for the experimental generation of a variety of vector beams, including vector Laguerre-Gauss (vLG) and vector Bessel-Gauss (vBG), and experimentally realize vector Mathieu-Gauss (vMG) beams for the first time, to the best of our knowledge. We require only a single binary hologram on a Digital Micromirror Device (DMD) and use two orthogonally polarized beams with complex conjugate amplitudes to obtain independent control over both the phase and polarization structure of the generated fields. We characterize the beams using intensity measurements and Stokes polarimetry, and quantify their vector quality through concurrence. The experimental results show excellent agreement with simulations, confirming that this setup can reliably produce high-quality vector beams. The approach is compact, cost-effective, and easily adaptable, making it well-suited for a wide range of applications in beam shaping and structured light.</p
TomoSAR from theory to practice:Overview of the advancements and challenges
No full textSynthetic Aperture Radar Tomography (TomoSAR) is a transformative technique for multi-dimensional imaging, enabling detailed analysis of complex environments. This paper provides an overview of TomoSAR, exploring its technological advancements and the practical challenges. We delve into the principles of TomoSAR and its ability to synthesize elevation arrays for precise scatterer separation in the vertical dimension. The paper reviews key advancements to enhance the robustness of the TomoSAR inversion process. We also discuss emerging solutions, particularly the application of Artificial Intelligence (AI) and deep learning models, as well as the ongoing development of spaceborne SAR systems with improved revisit times. Finally, the paper outlines future directions for TomoSAR, emphasizing its potential for global-scale monitoring and interdisciplinary collaboration to extend its capabilities in environmental and urban monitorin
Experimental and Theoretical Study of Frequency Combs in Hybrid Lasers with a Narrow-Band Mirror
No full textWe present experimental and theoretical evidence of a self-pulsing regime in III-V/SiN hybrid integrated lasers featuring a frequency-selective mirror. While such a regime has been previously theoretically predicted in microcavity laser, as in the case of Fano laser, our research demonstrates its occurrence in a simpler and more accessible silicon photonics platform. Our findings demonstrate that these lasers can generate narrow free spectral range (FSR) frequency combs, with FSR of just a few gigahertz and smaller than the cavity FSR. The experimental observations are also supported by a theoretical model.</p
Germany Ahr Valley Flood 2021: a social and mental health perspective of older adults (65+) experiences
No full textThis study investigates the social and mental health impacts of the 2021 Ahr Valley flood in Germany on older adults (aged 65+).We used a mixed-methods approach, combining household surveys, spatial GIS analysis, qualitative in-depth interviews, and group discussions with older adults, physicians, nurses, and social service providers. Mental health was assessed using the PHQ-4 and PC-PTSD-5, while social health was measured through adapted Likert scales on belonging, social support, and information access. Spatial analysis in ArcGIS Pro used self-reported flood depth data and secondary sources, applying inverse distance weighting and kernel density estimation to map perceived flooding and related health outcomes.Three years after the flood, 47% of older adults screened positive for anxiety, 38% for depression, and 39% for PTSD. Women reported slightly greater flood exposure than men. Social health scores were not statistically significant, but indicated strong place attachment and social support, and insufficient information flow during the disaster. Spatial analysis revealed clear hotspots of mental and social health challenges in severely affected areas like Bad Neuenahr-Ahrweiler and Dernau, and captured indoor flood water heights.For many, the Ahr Valley was not merely a geographic space but a social and emotional anchor that embodied their identity and life history. The study bridges environmental, social, and mental dimensions of disaster impact, providing evidence for more context-sensitive, community-based interventions. The findings contribute to European flood research by addressing knowledge gaps.We suggest a future need for longitudinal, interdisciplinary designs that capture older adults’ vulnerability and resilience after climate-related disasters
Digital Trace Data as Measurement Instruments for Variance-Theoretic Research in Information Systems
No full textDriven by the digitization of organizations, digital trace data offer novel insights into human behaviors with technology. Digital trace data are longitudinal records of tech-nology use. Over the last years, we have seen a surge in interest with growing empirical applications and research into the conceptual and methodological foundations of digital trace data research. So far, however, using digital trace data as a basis for measurement instruments in traditional variance-theoretical applications has received little attention, alt-hough they may enable novel analyses for theorizing from digitized contexts. The nascent research using digital trace data as measurement instruments has received critiques about validity problems, suggesting that guidelines for robust construct operationalizations are needed. Based on a literature review, this chapter identifies sources for validity problems with digital trace data. I further derive recommendations for assessing and reporting instru-ment validity with digital trace data. Thereby, this chapter contributes to improving the robustness of quantitative research using digital trace data
How does nanofiltration of (diluted or concentrated) sea water affect brine quality for the production of highly pure crystalline NaCl?
No full textNaCl salt can be produced from zero-liquid discharge sea water desalination, where nanofiltration is applied to increase brine purity. However, during the development of the nanofiltration process the focus is usually on reducing liquid waste discharge and increasing drinking water yield in reverse osmosis, rather than on producing salt of sufficient purity. This study shows that nanofiltration of (synthetic) sea water, or its marginally diluted or more concentrated derivatives, is able to strongly reduce divalent ion concentrations in brine. However, the obtained selectivity for potassium over sodium (approximately 1) and for bromide over chloride (increasing from 1 to 1.2–1.3, depending on the membrane type, as function of flux) lead to an increased monovalent impurity level in the purified brine as compared to brine obtained from solution mining, thereby requiring additional post treatment to produce high quality crystalline salt. Furthermore, obtained sodium and chloride retentions between 0 and 40 % reduce the NaCl concentration in the brine leading to the need for additional brine concentration to reach saturation prior to the crystallization process. DSPM-DE model results, using membrane characteristics determined from characterization and salt solution experiments, predict the obtained potassium over sodium selectivity properly, but underpredicts the obtained bromide over chloride selectivity with a predicted selectivity of approximately 1 for all experimental conditions and evaluated membranes. Ranking commercial membranes based on performance results obtained over a large variety of experimental conditions is not straightforward when permeate quality is considered as well
Sensor Fusion Using 1D-CNNs in Atrial Fibrillation Detection and Decision Support
No full textData fusion, which involves combining signals from multiple sensors simultaneously, enhances the accuracy of inferences drawn from complex datasets. This article presents a fusion method for multisensor data that utilizes 1-dimensional convolutional neural networks (1D-CNNs) to integrate the decisions of individual models detecting the same events across different physiological signals. The approach is designed to improve diagnostic accuracy by analyzing the degree of agreement between the decisions of these models. To further assist healthcare practitioners in their decision-making process, especially when discrepancies arise between the models, the fusion method incorporates explainable AI techniques, specifically the Local Interpretable Model-agnostic Explanations (LIME). These techniques provide insights into the model's decision-making process, making it easier for practitioners to trust the results. The proposed fusion architecture is specifically evaluated for atrial fibrillation detection using two multimodal signal inputs from a subset of the MIMIC III database. In this case, the fusion algorithm employs 1D-CNN-based models to detect atrial fibrillation from electrocardiogram (ECG) and photoplethysmogram (PPG) signals. The individual ECG and PPG models achieved high accuracies of 98.85% and 98.52%, with sensitivities of 100% and 99.39%, respectively. When the models were fused based on their agreement, the output demonstrated even better performance, achieving an accuracy of 99.85% and sensitivity of 99.80%. This approach underscores the potential of CNN-based fusion techniques in enhancing the reliability of physiological event detection
Mapping indicators for morphological informality in Nairobi, Kenya using satellite imagery
No full textOver one billion people globally live in slums, informal settlements, and other deprived areas. However, maps of deprived areas are often unavailable or over-simplistic, distinguishing only between slums and formal areas. Recent research advocates for a multidimensional approach to better account for the complexity of deprivation. Previous studies have mapped the unplanned urbanization domain of deprivation using morphometrics derived from building footprint data. This study explores leveraging Earth observation data for scalability and regular updates by using high-resolution satellite imagery and deep learning to map indicators for morphological informality in Nairobi, Kenya. The proposed model combines a ResNet backbone with three classification heads to map the two indicators: irregular settlement layout (ISL) and small, dense structures (SDS), alongside building presence. The model was trained on automatically generated reference data using building footprint morphometrics and clustering, and its outputs were validated through community-sourced annotations obtained via participatory action research. The results demonstrate the potential of high-resolution satellite imagery for mapping ISL (F1 80.90) and SDS (F1 78.73). Nonetheless, further research is required on the geographic transferability of the proposed method