Multiscale Character Research Articles

Stone inscription image segmentation based on Stacked-UNets and GANs

To overcome the challenges posed in effectively extracting stone inscriptions characterized by highly self-similarity between the foreground and background, a character image segmentation framework is proposed that integrates Stacked-UNets and Generative Adversarial Networks (GANs). Initially, a convolutional rule tailored for self-similar feature extraction is introduced to enhance the image detail segmentation. Subsequently, in improving the stacked units, multi-scale character masks with Spatial Transformer Network (STN) are added to guide the character segmentation. Finally, with two GANs and datasets, the character recognition and generation capabilities of the Stacked-UNets are trained, and further constructing its character segmentation and restoration abilities. Ultimately, the edge segmentation performance was enhanced, both in extracting characters from stone inscriptions with self-similar blocks and in effectively recovering partially incomplete characters. Compared to state-of-the-art methods, the Stacked-UNets demonstrates improvements of 6.41%, 7.39%, 8.43%, 8.11%, 4.13%, and 2.58% across six indicators, respectively.

A uniformly convergent method for two-parameter singularly perturbed parabolic partial differential equations with a large shift

A class of singularly perturbed two-parameter parabolic partial differential equations with a large shift is studied. These equations include time-dependent variables and exhibit sensitivity to small perturbations in the system parameters. Moreover, the shift captures the influence of neighbouring states or events on the current evolution of the system. It adds memory-like behaviour to the problem, making their analysis and numerical solution more intricate. The solution of these equations exhibits a multiscale character. The interior and boundary layers exist across which the solution has a steep gradient. A higher-order accurate numerical method is proposed on a non-uniform mesh to solve the problem. The method combines an upwind difference scheme in space and a Crank-Nicolson scheme in time. The method is unconditionally stable, and parameters uniformly convergent with second-order accuracy in time and first-order accuracy in the space variable. Besides, the paper presents numerical results and illustrations to support theoretical estimates.

Effective thermal conductivity of granular soils: a review of influencing factors and prediction models towards an investigation framework through multiscale characters

The effective thermal conductivity of soil is important to geo-engineering applications, and it is controlled by factors across different length scales. Through a comprehensive review of these factors, we found that while other more traditional factors have been well studied, there is still a lack of characterisation of soil microscale and mesoscale structures and their influence on effective thermal conductivity. In addition, after reviewing the models available in the literature for soil effective thermal conductivity prediction, it was found that compared with empirical and theoretical models, machine learning models can account for the influence of multi-scale factors; however, research into them is scarce. To overcome the limitations of previous research, we proposed a framework that can investigate the factors influencing soil effective thermal conductivity at multiple scales. It includes the impact of soil structural factors at micro to mesoscale, and this impact is integrated with the influence from other factors for accurate thermal conductivity prediction.

On the (Apparently) Paradoxical Role of Noise in the Recognition of Signal Character of Minor Principal Components

The usual method of separating signal and noise principal components on the sole basis of their eigenvalues has evident drawbacks when semantically relevant information ‘hides’ in minor components, explaining a very small part of the total variance. This situation is common in biomedical experimentation when PCA is used for hypothesis generation: the multi-scale character of biological regulation typically generates a main mode explaining the major part of variance (size component), squashing potentially interesting (shape) components into the noise floor. These minor components should be erroneously discarded as noisy by the usual selection methods. Here, we propose a computational method, tailored for the chemical concept of ‘titration’, allowing for the unsupervised recognition of the potential signal character of minor components by the analysis of the presence of a negative linear relation between added noise and component invariance.

Nomes, lugares e malocas de dança: unidades sociais entre os Tukano orientais

Abstract Since Irving Goldman’s pioneering work among the Cubeo (Goldman, 1968 [1963]), defining Eastern Tukanoan social units has been a recurring issue in regional ethnology. Many late twentieth-century ethnographies produced solid knowledge on Tukano social structure by showing distinctive and exclusive features within the Amazonian scenario (such as language exogamy, patrilineality, specialized hierarchical clans or sibs, etc.). However, they also indicated some inconsistencies and paradoxes related to a discrepancy between the ideal model of hierarchy and patrilineality and a dynamic reality more attuned to equality and cognatic arrangements. Recent attempts to understand these paradoxes have focused on what a social unit actually means from an indigenous perspective, leading to new ideas about the fractal (multi-scale) character of such units and the way they work. These interpretations take the maloca (communal house) as a model, which evolves as it changes names, places, rituals, and specific narratives that serve to create permanent differences, which in turn produce disputes and alter kinship relations.

Complex structure of turbulence across the ASDEX Upgrade pedestal

The theoretical investigation of relevant turbulent transport mechanisms in H-mode pedestals is a great scientific and numerical challenge. In this study, we address this challenge by global, nonlinear gyrokinetic simulations of a full pedestal up to the separatrix, supported by a detailed characterisation of gyrokinetic instabilities from just inside the pedestal top to the pedestal centre and foot. We present ASDEX Upgrade pedestal simulations using an upgraded version of the gyrokinetic, Eulerian, delta-f code GENE (genecode.org) that enables stable global simulations at experimental plasma $\beta$ values. The turbulent transport is found to exhibit a multi-channel, multi-scale character throughout the pedestal with the dominant contribution transitioning from ion-scale trapped electron modes/micro-tearing modes at the pedestal top to electron-scale electron temperature gradient modes in the steep gradient region. Consequently, the turbulent electron heat flux changes from ion to electron scales and the ion heat flux reduces to almost neoclassic values in the pedestal centre. $E\times B$ shear is found to strongly reduce heat flux levels in all channels (electron, ion, electrostatic, electromagnetic) and the interplay of magnetic shear and pressure gradient is found to locally stabilise ion-scale instabilities.

U-Net Architecture for Ancient Handwritten Chinese Character Detection in Han Dynasty Wooden Slips

Recent character detectors have been modeled using deep neural networks and have achieved high performance in various tasks, such as text detection in natural scenes and character detection in historical documents. However, existing methods cannot achieve high detection accuracy for wooden slips because of their multi-scale character sizes and aspect ratios, high character density, and close character-to-character distance. In this study, we propose a new U-Net-based character detection and localization framework that learns character regions and boundaries between characters. The proposed method enhances the learning performance of character regions by simultaneously learning the vertical and horizontal boundaries between characters. Furthermore, by adding simple and low-cost post-processing using the learned regions of character boundaries, it is possible to more accurately detect the location of a group of characters in a close neighborhood. In this study, we construct a wooden slip dataset. Experiments demonstrated that the proposed method outperformed existing character detection methods, including state-of-the-art character detection methods for historical documents.

On the non-self-adjoint and multiscale character of passive scalar mixing under laminar advection

Except in the trivial case of spatially uniform flow, the advection–diffusion operator of a passive scalar tracer is linear and non-self-adjoint. In this study, we exploit the linearity of the governing equation and present an analytical eigenfunction approach for computing solutions to the advection–diffusion equation in two dimensions given arbitrary initial conditions, and when the advecting flow field at any given time is a plane parallel shear flow. Our analysis illuminates the specific role that the non-self-adjointness of the linear operator plays in the solution behaviour, and highlights the multiscale nature of the scalar mixing problem given the explicit dependence of the eigenvalue–eigenfunction pairs on a multiscale parameter $q=2{\rm i}k\,{\textit {Pe}}$ , where $k$ is the non-dimensional wavenumber of the tracer in the streamwise direction, and ${\textit {Pe}}$ is the Péclet number. We complement our theoretical discussion on the spectra of the operator by computing solutions and analysing the effect of shear flow width on the scale-dependent scalar decay of tracer variance, and characterize the distinct self-similar dispersive processes that arise from the shear flow dispersion of an arbitrarily compact tracer concentration. Finally, we discuss limitations of the present approach and future directions.

An FFT-based framework for predicting corrosion-driven damage in fractal porous media

Understanding fracture in cementitious materials caused by the deposition and growth of corrosion products requires scale-bridging approaches due to the large length-scale difference between the micro-pores, where deposition occurs, and the structure, where deterioration manifests. Cementitious materials bear a highly heterogeneous micro-structure owing to the fractal nature of micro-pores. Simultaneously, a corrosion-driven fracture is a multi-physics problem involving ionic diffusion, chemical reactions, and stress development. This multi-scale and multi-physical character makes scale-bridging studies computationally costly, often leading to the use of simplified fractal porous media, which has important consequences for the quantitative interpretation of the results. Recent advances in homogenization approaches using Fast-Fourier-Transform (FFT) based methods have raised interest due to their ease of implementation and low computational cost. This paper presents an FFT-based framework for solving corrosion-driven fractures within fractal porous media. We demonstrate the effectiveness of the Fourier-based spectral method in resolving the multiple corrosion-driven mechanisms such as ionic diffusion, stress development, and damage within a fractal porous microstructure. Based on the presented methodology, we analyse the impact of simplifying fractal porous media with simple Euclidean geometry on corrosion-driven fracture. Our results demonstrate the importance of preserving both the porosity and fractal nature of pores for precise and reliable modelling of corrosion-driven failure mechanisms.

Lightweight Spatial-Spectral Network Based on 3D-2D Multi-Group Feature Extraction Module for Hyperspectral Image Classification

ABSTRACT Each pixel in the hyperspectral image contains tens to hundreds of narrow bands of detailed spectral information captured by a hyperspectral sensor. Hyperspectral image classification (HSIC) is widely used in remote sensing image analysis. Convolutional neural network (CNN) is one of the most commonly used deep learning methods based on visual data processing. In recent years, hyperspectral classification method based on the convolutional neural network has shown excellent performance. However, single-scale feature extraction may lose some important detail information and cannot guarantee that the best spatial features are captured. Multi-scale character extraction and increasingly deep network structures can improve classification accuracy, but also bring a large number of computational parameters, resulting in higher computational costs. In response to this series of problems, this paper proposes a lightweight spatial-spectral network based on 3D − 2D multi-group feature extraction module (MGFM) for HSCI. In the 3D-MGFM, the input feature maps are grouped and processed in parallel, and then the information of each channel is integrated by point-wise convolution to achieve spatial-spectral feature fusion. We use dilated convolutions with different dilation rate on the convolution kernel to improve the problem of feature information loss, so that we can better learn deep-level spatial-spectral features. In the 2D-MGFM, we selectively emphasize spectral feature information through SENet and extract multi-scale features based on Depthwise Separable Convolution (DSC) grouping. Experimental results on four hyperspectral datasets demonstrate that the proposed method achieves better classification performance than other state-of-the-art methods.

Three-scale bridging for woven composites using homogenization techniques

Woven composites have a multiscale character, whose structural response is affected by mechanisms on different lower scales. The predicted macroscopic response of a carbon fiber reinforced woven composite is achieved by model order reduction and homogenization techniques, taking into account the structures and mechanisms on both mesoscopic and microscopic scales. Reduced order models based on piecewise uniform fields and mean field homogenization are integrated for the micro-meso-macro upscaling to finally determine the effective macroscopic response of a woven unit cell. A spatial decomposition for the woven composite unit cell on the mesoscale is implemented, achieving the model order reduction by taking into account the local microstructure of the yarns as well as inelastic fields emerging under selected deformation conditions. Following, different numerical tests are performed on the woven unit cell, containing complex and inelastic loading histories. Macroscopic homogenized predictions following the reduced order models and full-field direct numerical simulations are compared. Very high accuracies are achieved by the employed reduced order models, allowing for great computational savings.

Handwritten Multi-Scale Chinese Character Detector with Blended Region Attention Features and Light-Weighted Learning.

Character-level detection in historical manuscripts is one of the challenging and valuable tasks in the computer vision field, related directly and effectively to the recognition task. Most of the existing techniques, though promising, seem not powerful and insufficiently accurate to locate characters precisely. In this paper, we present a novel algorithm called free-candidate multiscale Chinese character detection FC-MSCCD, which is based on lateral and fusion connections between multiple feature layers, to successfully predict Chinese characters of different sizes more accurately in old documents. Moreover, cheap training is exploited using cheaper parameters by incorporating a free-candidate detection technique. A bottom-up architecture with connections and concatenations between various dimension feature maps is employed to attain high-quality information that satisfies the positioning criteria of characters, and the implementation of a proposal-free algorithm presents a computation-friendly model. Owing to a lack of handwritten Chinese character datasets from old documents, experiments on newly collected benchmark train and validate FC-MSCCD to show that the proposed detection approach outperforms roughly all other SOTA detection algorithms.

An open source framework for advanced Multi-physics and multiscale modelling of solid oxide fuel cells

Solid oxide fuel cells are high-efficiency renewable energy devices and considered one of the most promising net-zero carbon energy technologies. Numerical modelling is a powerful tool for the virtual design and optimisation of the next-generation solid oxide fuel cells but needs to tackle issues for incorporating the multi-scale character of the cell and further improving the accuracy and computational efficiency. While most of solid oxide fuel cell models were developed based on closed source platforms which limit the freedom of customisation in numerical discretization schemes and community participation. Here, an open source multi-physics and multiscale platform for advanced SOFC simulations consisting of both cell- and pore-scale performance models was developed using OpenFOAM. The modelling aspects are elucidated in detail, involving the coupling of various physical equations and the implementation of the pore-scale electrode in the performance model. The entire platform was carefully validated against experimental data and the other numerical models which were implemented in commercial software ANSYS Fluent and based on the lattice Boltzmann method. The cell-scale model is subsequently employed to study the effects of different fuels, i.e., pure hydrogen and different ratios of pre-reformed methane gas under various operating temperatures. It is found that the cell-scale model reasonably predicts the effects of these parameters on the cell performance, aligning well with the Fluent model. This study further identified the size of representative element volume with respect to the current density for the anode via the pore-scale model where the realistic microstructures reconstructed by a Xe plasma focused ion beam–scanning electron microscopy are employed as computational domains. It is found that a volume element size of 1243 voxels is sufficient to yield the representative current density of the whole. All these numerical investigations show that OpenFOAM is a potential multi-physics and multi-scale computational platform that is capable of accurately predicting both cell-scale and pore-scale performance and spatial information of solid oxide fuel cells. The developed models are also made public in GitHub to inspire community to further develop around it.

Multi-scale temporal characters mining for bird activities based on historical avian radar system datasets

AbstractAvian radar systems are effective for wide-area bird detection and tracking, but application significances need further exploration. Existing radar data mining methods provide long-term functionalities, but they are problematic for bird activity modelling especially in temporal domain. This paper complements this insufficiency by introducing a temporal bird activity extraction and interpretation method. The bird behaviour is quantified as the activity degree which integrates intensity and uncertainty characters with an entropy weighing algorithm. The method is applicable in multiple temporal scales. Historical radar dataset from a system deployed in an airport is adopted for verification. Temporal characters demonstrate good consistency with understandings from local observers and ornithologists. Daily commuting and roosting characters of local birds are well reflected, evening bat activities are also extracted. Night migration activities are demonstrated clearly. Results indicate the proposed method is effective in temporal bird activity modelling and interpretation. Its integration with bird strike risk models might be more useful for airport safety management with wildlife interference.

A higher-order uniformly convergent defect correction method for singularly perturbed convection-diffusion problems on an adaptive mesh

This paper presents a defect correction scheme based on finite-difference discretization for a singularly perturbed convection-dominated diffusion problem. The solution of this class of problems exhibits a multiscale character. There are narrow regions in which the solution grows exponentially and reveals layer behaviour. The defect correction method that we propose improves the efficiency of a numerical solution through iterative improvement and generates a stable higher-order method over an adaptive non-uniform polynomial-Shishkin mesh. An extensive theoretical analysis is presented, which establishes that the method is second-order uniformly convergent and highly stable. The convergence obtained is optimal because it is free from any logarithmic term. The numerical result for two model problems is presented, agreeing with the theoretical estimates. Furthermore, we compare the results with those of other non-uniform mesh found in the literature.

Mesh generation and multi-scale simulation of a contracting muscle–tendon complex

The multi-scale character of skeletal muscle models requires simulations with high spatial resolution to capture all relevant effects. This naturally involves high computational load that can only be tackled by parallel computations. We simulate electrophysiology and muscle contraction using a state-of-the-art, biophysical chemo-electro-mechanical model that requires meshes of the 3D domain with embedded, aligned 1D meshes for muscle fibers. We present novel algorithms to construct highly-resolved meshes with robust properties for real muscle geometries from surface triangulations. We demonstrate their use and suitability in a simulation of the biceps brachii muscle and tendons. In addition, the respective simulations showcase several functional enhancements of our simulation framework OpenDiHu.

FAN-MCCD: Fast and Accurate Network for Multi-Scale Chinese Character Detection.

Inaccurate localization due to scale-variation during character detection causes a widespread issue overconfidence in results of the document analysis community, for the most part in historical and handwritten documents. In this work, we explored the performance of a state-of-the-art network with a simple pipeline that fast and accurately predicts handwritten Chinese characters in old documents. In order to adapt to locations of characters with multi-scale more precisely, excluding pre-processing and in-between steps, we utilized a network with multi-scale feature maps. Then, across each feature map, pre-selected boxes of unalike scales and aspect ratios were employed. The last step was to prune the bounding boxes, sending them to non-maximum suppression to yield the final results. Focusing on a well-designed neural network architecture and loss function that presents well-classified examples, we found our experiments on Caoshu, Character, and Src-images datasets demonstrated that detection performance was enhanced for the detection rate (DT), the false positive per character (FPPC), and the F-score in the order of 98.84%, 0.71, and 97.64%, respectively. In comparison with SSD (single-shot detector), the detection performance of a detection rate (DT), the false positive per character (FPPC), and the F-score were 61.12%, 6.12, and 60.33%, respectively.

Some mathematical tools for modelling malaria: a subjective survey

In this paper, we provide a brief survey of mathematical modelling of malaria and how it is used to understand the transmission and progression of the disease and design strategies for its control to support public health interventions and decision-making. We discuss some of the past and present contributions of mathematical modelling of malaria, including the recent development of modelling the transmission-blocking drugs. We also comment on the complexity of the malaria dynamics and, in particular, on its multiscale character with its challenges and opportunities. We illustrate the discussion by presenting a curve fitting using a 95% confidence interval for the South African data for malaria from the years 2001-2018$ and provide projections for the number of malaria cases and deaths up to the year 2025.

Theory of delayed response in river morphodynamics: Applicability and limitations

The delayed response of fluvial rivers to external disturbances has been described by many researchers. To simulate such behavior, the rate law model (or the delayed response model) was developed by previous researchers, and has been applied to a series of river morphological problems. However, to date, the applicability of the rate law model has not been fully understood. In the current paper, a physically-based analysis of the rate law model is presented to assess the responses of bed elevation and the grain size of surface sediment, using the response of the Shi-ting River, China, after the 2008 Wenchuan Earthquake as an example. First, a physically-based river morphodynamic model is implemented to reproduce the post-earthquake adjustments of the Shi-ting River. Next, the mathematical properties of the Hirano–Exner equation, which describes the dynamics of both bed elevation and surface texture, are analyzed. It is shown that the dynamics of bed elevation are dominated by a diffusion process and the analytical solution for bed elevation has a similar mathematical formulation to that of the rate law model. Thus, the rate law and morphodynamic models similarly predict the adjustment of bed elevations. In contrast, the dynamics of bed surface texture are controlled by both advection and diffusion processes. The advection and diffusion processes dominate adjustments over the short (annual to decadal scale) and long (century to millenial scale) terms, respectively. These physics produce the multi-scale and non-monotonic character of the adjustment of bed surface texture. As a result, the rate law model is incapable of describing the adjustment of bed surface texture. These findings highlight the applicability and limitations of the rate law model in simulating river morphodynamic processes.

Multiscale criticality measures as general-purpose gauges of proper brain function

The brain is universally regarded as a system for processing information. If so, any behavioral or cognitive dysfunction should lend itself to depiction in terms of information processing deficiencies. Information is characterized by recursive, hierarchical complexity. The brain accommodates this complexity by a hierarchy of large/slow and small/fast spatiotemporal loops of activity. Thus, successful information processing hinges upon tightly regulating the spatiotemporal makeup of activity, to optimally match the underlying multiscale delay structure of such hierarchical networks. Reduced capacity for information processing will then be expressed as deviance from this requisite multiscale character of spatiotemporal activity. This deviance is captured by a general family of multiscale criticality measures (MsCr). MsCr measures reflect the behavior of conventional criticality measures (such as the branching parameter) across temporal scale. We applied MsCr to MEG and EEG data in several telling degraded information processing scenarios. Consistently with our previous modeling work, MsCr measures systematically varied with information processing capacity: MsCr fingerprints showed deviance in the four states of compromised information processing examined in this study, disorders of consciousness, mild cognitive impairment, schizophrenia and even during pre-ictal activity. MsCr measures might thus be able to serve as general gauges of information processing capacity and, therefore, as normative measures of brain health.