Multiscale Implementation Research Articles

Design and Implementation of Novel Hybrid and Multiscale- Assisted CNN and ResNet Using Heuristic Advancement of Adaptive Deep Segmentation for Iris Recognition

Due to its significant applications in security, the iris recognition process has been considered as the most active research area over the last few decades. In general, the iris recognition framework has been crucially utilized for various security applications because it includes a set of features as well as does not alter its character according to the time. In recent times, emerging deep learning techniques have attained huge success, particularly in the field of the iris recognition framework model. Moreover, in considering the field of iris recognition, there is no possibility for the remarkable capability of the deep learning model as well as to attain superior performance. To handle the issues in the conventional model of iris recognition, a novel heuristic-aided deep learning framework has been implemented for recognizing the iris system. Initially, the required source iris images are gathered from the data sources. It is then followed by the pre-processing stage, where the pre-processed image is obtained. Consequently, the image segmentation process is carried out by Adaptive Deeplabv3+layers, in which the parameters are optimized using the Modified Weighted Flow Direction Algorithm (MWFDA). Finally, the iris recognition is accomplished by hybrid Hybridization of Multiscale Dilated-Assisted Learning (MDAL) that will be composed of both a Convolutional Neural Network (CNN) and a Residual Network (ResNet). To achieve optimal recognition results, the parameters in CNN and ResNet are tuned optimally by using MWFDA. The experimental results are estimated with the help of distinct measures. Contrary to conventional methods, the empirical results prove that the recommended model achieves the desired value to enhance the recognition performance.

Evaluating the Multiscale Implementation of Valid Time Shifting within a Real-Time EnVar Data Assimilation and Forecast System for the 2022 HWT Spring Forecasting Experiment

Abstract Multiscale valid time shifting (VTS) was explored for a real-time convection-allowing ensemble (CAE) data assimilation (DA) system featuring hourly assimilation of conventional in situ and radar reflectivity observations, developed by the Multiscale Data Assimilation and Predictability Laboratory. VTS triples the base ensemble size using two subensembles containing member forecast output before and after the analysis time. Three configurations were tested with 108-member VTS-expanded ensembles: VTS for individual mesoscale conventional DA (ConVTS) or storm-scale radar DA (RadVTS) and VTS integrated to both DA components (BothVTS). Systematic verification demonstrated that BothVTS matched the DA spread and accuracy of the best-performing individual component VTS. The 10-member forecasts showed BothVTS performs similarly to ConVTS, with RadVTS having better skill in 1-h precipitation at forecast hours 1–6, while Both/ConVTS had better skill at later hours 7–15. An objective splitting of cases by 2-m temperature cold bias revealed RadVTS was more skillful than Both/ConVTS out to hour 10 for cold-biased cases, while BothVTS performed best at most hours for less-biased cases. A sensitivity experiment demonstrated improved performance of BothVTS when reducing the underlying model cold bias. Diagnostics revealed enhanced spurious convection of BothVTS for cold-biased cases was tied to larger analysis increments in temperature than moisture, resulting in erroneously high convective instability. This study is the first to examine the benefits of a multiscale VTS implementation, showing that BothVTS can be utilized to improve the overall performance of a multiscale CAE system. Further, these results underscore the need to limit biases within a DA and forecast system to best take advantage of VTS analysis benefits.

BULLDOZER: AN AUTOMATIC SELF-DRIVEN LARGE SCALE DTM EXTRACTION METHOD FROM DIGITAL SURFACE MODEL

Abstract. This study presents a Digital Terrain Model extraction method called Bulldozer. The only required input of Bulldozer is a Digital Surface Model generated from any sensors (usually optical or LIDAR) with any kind of software. After reviewing both the initial DrapCloth algorithm (Zhang et al., 2016) and its multi scale implementation (Leotta et al., 2019), some issues have been highlighted when extracting DTM from stereo satellite images such as the lost of ground adhesion under rising terrain areas, the appearance of sinks due to correlation issues when computing the DSM and finally the lack of scalability when processing large input data. Bulldozer has been developed to tackle all these issues and proposes a full automatic scalable pipeline composed of a pre-processing step to clean noisy DSMs by detecting and smoothing disturbed areas, a DTM extraction step based on a modified DrapCloth algorithm to stick to the ground under rising terrain and a post-processing step to smooth sharp sinks. The scalability has been solved using a tiling strategy and the definition of a stability margin that ensures identical results to those obtained if the whole DSM would have been processed at once in memory. As a result, Bulldozer outperforms its concurrent with respect to runtime execution while providing high quality DTMs over various types of landscapes.

Towards HPC simulations of billion-cell reservoirs by multiscale mixed methods

A three dimensional parallel implementation of Multiscale Mixed Methods based on non-overlapping domain decomposition techniques is proposed for multi-core computers and its computational performance is assessed by means of numerical experiments. As a prototypical method, from which many others can be derived, the Multiscale Robin Coupled Method is chosen and its implementation explained in detail. Numerical results for problems ranging from millions up to more than 2 billion computational cells in highly heterogeneous anisotropic rock formations based on the SPE10 benchmark are shown. The proposed implementation relies on direct solvers for both local problems and the interface coupling system. We find good weak and strong scalalability as compared against a state-of-the-art global fine grid solver based on Algebraic Multigrid preconditioning in single and two-phase flow problems.

Variational image motion estimation by preconditioned dual optimization

<p style='text-indent:20px;'>Estimating optical flows is one of the most interesting problems in computer vision, which estimates the essential information about pixel-wise displacements between two consecutive images. This work introduces an efficient dual optimization framework with accelerated preconditioners to the challenging nonsmooth optimization problem of total-variation regularized optical-flow estimation. In theory, the proposed dual optimization framework brings an elegant variational analysis to the given difficult optimization problem, while presenting an efficient algorithmic scheme without directly tackling the corresponding nonsmoothness in numeric. By introducing efficient preconditioners with a multi-scale implementation, the proposed preconditioned dual optimization approaches achieve competitive estimation results of image motion, compared to the state-of-the-art methods. Moreover, we show that the proposed preconditioners can guarantee convergence of the implemented numerical schemes with high efficiency.</p>

Implementation of VBM3D and Some Variants

VBM3D is an extension to video of the well-known image denoising algorithm BM3D, which takes advantage of the sparse representation of stacks of similar patches in a transform domain. The extension is rather straightforward: the similar 2D patches are taken from a spatio-temporal neighborhood which includes neighboring frames. In spite of its simplicity, the algorithm offers a good trade-off between denoising performance and computational complexity. In this work we revisit this method, providing an open-source C++ implementation reproducing the results. A detailed description is given and the choice of parameters is thoroughly discussed. Furthermore, we study and compare several extensions of the original algorithm: (1) a multiscale implementation, (2) the use of 3D patches, (3) the use of optical flow to guide the patch search. These extensions yield results which are competitive with the most recent state of the art.

CycleMorph: Cycle consistent unsupervised deformable image registration.

Image registration is a fundamental task in medical image analysis. Recently, many deep learning based image registration methods have been extensively investigated due to their comparable performance with the state-of-the-art classical approaches despite the ultra-fast computational time. However, the existing deep learning methods still have limitations in the preservation of original topology during the deformation with registration vector fields. To address this issues, here we present a cycle-consistent deformable image registration, dubbed CycleMorph. The cycle consistency enhances image registration performance by providing an implicit regularization to preserve topology during the deformation. The proposed method is so flexible that it can be applied for both 2D and 3D registration problems for various applications, and can be easily extended to multi-scale implementation to deal with the memory issues in large volume registration. Experimental results on various datasets from medical and non-medical applications demonstrate that the proposed method provides effective and accurate registration on diverse image pairs within a few seconds. Qualitative and quantitative evaluations on deformation fields also verify the effectiveness of the cycle consistency of the proposed method.

Paris Climate Agreement: Promoting Interdisciplinary Science and Stakeholders’ Approaches for Multi-Scale Implementation of Continental Carbon Sequestration

The Paris Climate Agreements and Sustainable Development Goals, signed by 197 countries, present agendas and address key issues for implementing multi-scale responses for sustainable development under climate change—an effort that must involve local, regional, national, and supra-national stakeholders. In that regard, Continental Carbon Sequestration (CoCS) and conservation of carbon sinks are recognized increasingly as having potentially important roles in mitigating climate change and adapting to it. Making that potential a reality will require indicators of success for various stakeholders from multidisciplinary backgrounds, plus promotion of long-term implementation of strategic action towards civil society (e.g., law and policy makers, economists, and farmers). To help meet those challenges, this discussion paper summarizes the state of the art and uncertainties regarding CoCS, taking an interdisciplinary, holistic approach toward understanding these complex issues. The first part of the paper discusses the carbon cycle’s bio-geophysical processes, while the second introduces the plurality of geographical scales to be addressed when dealing with landscape management for CoCS. The third part addresses systemic viability, vulnerability, and resilience in CoCS practices, before concluding with the need to develop inter-disciplinarity in sustainable science, participative research, and the societal implications of sustainable CoCS actions.

Fast infrared and visible image fusion with structural decomposition

Infrared and visible image fusion aims to generate a composite image with salient thermal targets and texture information from infrared image and visible image. Existing advanced methods tend to consume high computational cost for a high-quality fusion result. In this paper, a simple yet effective method is proposed based on structural patch decomposition. A modified gamma function is proposed to weight the mean intensity component to keep the thermal target. An enhanced power function is used to weight the mean-removed component to preserve the texture information. Different from general patch level fusion implemented via a sliding window, we convert the explicit structural patch decomposition and fusion into an image level mean filtering via a detailed analysis on input images. By this means, the computational cost of the proposed method can be largely reduced, which is independent of the patch size. Furthermore, we analyze the relationship of our method with classic filtering based image decomposition methods. Finally, a multi-scale implementation of the proposed method is developed to avoid the evident halo and spatial inconsistency artifacts. Extensive experimental results on the public dataset demonstrate that the proposed method can obtain more texture information and outperform the state-of-the-art fusion method qualitatively and quantitatively. The code can be downloaded from: https://github.com/xiaohuiben/MSID-KBS.

Multi-Scale Implementation of 3D-RISM to the Electronic Structure Theory Being Applicable for Solvated Biomolecules

Multi-Scale Implementation of 3D-RISM to the Electronic Structure Theory Being Applicable for Solvated Biomolecules

Mobility of dislocations in aluminum: The role of non-Schmid stress state

The extent that non-Schmid stresses influence the mobility of dislocations in Al is studied using molecular dynamics (MD) simulations. Specifically, mobility laws for straight screw, 30°, 60° and edge dislocations are atomistically derived for different combinations of Escaig stress, τe, and stress normal to the (111) slip plane, σn. MD simulations show that the mobility of each dislocation is distinctly influenced by non-Schmid stresses. Furthermore, MD simulations of dislocation shear loop expansion show that mobility laws derived from straight dislocations are applicable to describe the more complex behavior of stress-state dependent expansion of a dislocation loop. Data describing the dislocation mobility laws are incorporated into discrete dislocation dynamics (DDD) simulations using a hierarchical multiscale approach. DDD simulations of isolated dislocation loop expansion show strong agreement with MD simulation results, validating the nonlinear multiscale implementation. DDD simulations of dislocation network evolution show that the use of stress state dependent dislocation mobility laws provides quantifiable changes to the plastic deformation path, which leads to different final microstructures. The results presented here demonstrate the key role that local stresses, experienced by the dislocation core, play on the evolution of a dislocation network during plastic deformation.

Computationally Efficient Concurrent Multiscale Framework for the Linear Analysis of Composite Structures

This paper presents a novel multiscale framework based on higher-order one-dimensional finite element models. The refined finite element models originate from the Carrera unified formulation (CUF): a novel and efficient methodology to develop higher-order structural theories hierarchically via a variable kinematic approach. The concurrent multiscale framework consists of a macroscale model to describe the structural level components interfaced with efficient CUF micromechanical models. Such micromechanical models can take into account the detailed architecture of the microstructure with high fidelity. The framework derives its efficiency from the capability of CUF models to detect accurate three-dimensional-like stress fields at reduced computational costs. This paper also shows the ability of the framework to interface with different classes of representative volume elements and the benefits of parallel implementations. The numerical cases focus on composite and sandwich structures and demonstrate the high fidelity and feasibility of the proposed framework. The efficiency of the framework stems from comparisons with the analysis time and memory requirement against traditional multiscale implementations.

Industrial Applications of Enzymes: Recent Advances, Techniques, and Outlooks

Enzymes as industrial biocatalysts offer numerous advantages over traditional chemical processes with respect to sustainability and process efficiency. Enzyme catalysis has been scaled up for commercial processes in the pharmaceutical, food and beverage industries, although further enhancements in stability and biocatalyst functionality are required for optimal biocatalytic processes in the energy sector for biofuel production and in natural gas conversion. The technical barriers associated with the implementation of immobilized enzymes suggest that a multidisciplinary approach is necessary for the development of immobilized biocatalysts applicable in such industrial-scale processes. Specifically, the overlap of technical expertise in enzyme immobilization, protein and process engineering will define the next generation of immobilized biocatalysts and the successful scale-up of their induced processes. This review discusses how biocatalysis has been successfully deployed, how enzyme immobilization can improve industrial processes, as well as focuses on the analysis tools critical for the multi-scale implementation of enzyme immobilization for increased product yield at maximum market profitability and minimum logistical burden on the environment and user.

Effects of precipitates and dislocation loops on the yield stress of irradiated iron

Plastic deformation of crystalline materials is governed by the features of stress-driven motion of dislocations. In the case of irradiated steels subject to applied stresses, small dislocation loops as well as precipitates are known to interfere with the dislocation motion, leading to an increased yield stress as compared to pure crystals. We study the combined effect of precipitates and interstitial glissile frac{{bf{1}}}{{bf{2}}}langle {bf{111}}rangle dislocation loops on the yield stress of iron, using large-scale three-dimensional discrete dislocation dynamics simulations. Precipitates are included in the simulations using our recent multi-scale implementation [A. Lehtinen et al., Phys. Rev. E 93 (2016) 013309], where the strengths and pinning mechanisms of the precipitates are determined from molecular dynamics simulations. In the simulations we observe dislocations overcoming precipitates with an atypical Orowan mechanism which results from pencil-glide of screw segments in iron. Even if the interaction mechanisms with dislocations are quite different, our results suggest that in relative terms, precipitates and loops of similar sizes contribute equally to the yield stress in multi-slip conditions.

Goal-oriented adaptivity of mixed GMsFEM for flows in heterogeneous media

We present an adaptive goal-oriented framework with the mixed Generalized Multiscale Finite Element Method (GMsFEM) to numerically solve the flow problem in heterogeneous media. In this paper, three different error indicators are considered: two, goal-oriented, and one a standard adaptive method for the global error. Each is used to approximate a number of quantities of interest given by interior edge fluxes in a high-contrast field. Following the setting of mixed GMsFEM, on each coarse neighborhood of the computational domain, the local error indicator introduced here is computed by the product of the residual norms of the primal and dual variational equations over coarse neighborhoods. Numerical results demonstrate the efficiency of this goal-oriented method in comparison to a multiscale implementation of the dual weighted residual method, and a standard adaptive method.

Reducing the computational requirements for simulating tunnel fires by combining multiscale modelling and multiple processor calculation

Multiscale modelling of tunnel fires that uses a coupled 3D (fire area) and 1D (the rest of the tunnel) model is seen as the solution to the numerical problem of the large domains associated with long tunnels. The present study demonstrates the feasibility of the implementation of this method in FDS version 6.0, a widely used fire-specific, open source CFD software. Furthermore, it compares the reduction in simulation time given by multiscale modelling with the one given by the use of multiple processor calculation. This was done using a 1200m long tunnel with a rectangular cross-section as a demonstration case. The multiscale implementation consisted of placing a 30MW fire in the centre of a 400m long 3D domain, along with two 400m long 1D ducts on each side of it, that were again bounded by two nodes each. A fixed volume flow was defined in the upstream duct and the two models were coupled directly. The feasibility analysis showed a difference of only 2% in temperature results from the published reference work that was performed with Ansys Fluent (Colella et al., 2010). The reduction in simulation time was significantly larger when using multiscale modelling than when performing multiple processor calculation (97% faster when using a single mesh and multiscale modelling; only 46% faster when using the full tunnel and multiple meshes). In summary, it was found that multiscale modelling with FDS v.6.0 is feasible, and the combination of multiple meshes and multiscale modelling was established as the most efficient method for reduction of the calculation times while still maintaining accurate results. Still, some unphysical flow oscillations were predicted by FDS v.6.0 and such results must be treated carefully.

Multiscale implementation of infinite-swap replica exchange molecular dynamics

Replica exchange molecular dynamics (REMD) is a popular method to accelerate conformational sampling of complex molecular systems. The idea is to run several replicas of the system in parallel at different temperatures that are swapped periodically. These swaps are typically attempted every few MD steps and accepted or rejected according to a Metropolis-Hastings criterion. This guarantees that the joint distribution of the composite system of replicas is the normalized sum of the symmetrized product of the canonical distributions of these replicas at the different temperatures. Here we propose a different implementation of REMD in which (i) the swaps obey a continuous-time Markov jump process implemented via Gillespie's stochastic simulation algorithm (SSA), which also samples exactly the aforementioned joint distribution and has the advantage of being rejection free, and (ii) this REMD-SSA is combined with the heterogeneous multiscale method to accelerate the rate of the swaps and reach the so-called infinite-swap limit that is known to optimize sampling efficiency. The method is easy to implement and can be trivially parallelized. Here we illustrate its accuracy and efficiency on the examples of alanine dipeptide in vacuum and C-terminal β-hairpin of protein G in explicit solvent. In this latter example, our results indicate that the landscape of the protein is a triple funnel with two folded structures and one misfolded structure that are stabilized by H-bonds.

Adaptation and Evaluation of an Optical Flow Method Applied to Coregistration of Forest Remote Sensing Images

The coregistration of heterogeneous geospatial images is useful in various remote sensing applications. Since the number of available data increases and the resolution improves, it is interesting to have an approach as automated, fast, robust, and accurate as possible. In this paper, we present a solution based on optical-flow computation. This algorithm called GeFolki allows the registration of images in a nonparametric and dense way. GeFolki is based on a local method of optical flow derived from the Lucas–Kanade algorithm, with a multiscale implementation, and a specific filtering including rank filtering, rolling guidance filtering and local contrast inversion. The efficiency of our coregistration chain is shown on radar, LIDAR, and optical images on Remningstorp forest in Sweden. An analysis of the relevant parameters is investigated for several scenarios. Finally, we demonstrate the accuracy of our coregistration by proposing specific metrics for LIDAR/radar coregistration, and optics/radar coregistration.

Enhanced composite damping through engineered interfaces

The damping properties of unidirectional, laminated, and woven composites have been predicted using a multiscale implementation of the High-Fidelity Generalized Method of Cells micromechanics theory. This model considers periodic repeating unit cell geometries on both the global and local scales and utilizes the constituent material specific damping coefficients, mechanical properties, and local fields, along with the strain energy approach, to determine effective directional specific damping coefficients of the composite. In addition to comparisons of the HFGMC predictions with results from the literature, the effect of a degraded fiber/matrix interface was examined parametrically. A significant finding was that strong maxima exist in the predicted composite damping coefficients as a function of degree of interfacial mechanical degradation. This suggests that drastic improvements in damping in composites can be achieved by properly engineering the fiber/matrix interface. The multiscale HFGMC simulations presented illustrate that the decrease in composite mechanical properties caused by such an engineered interface can be minimized when implemented within a technologically relevant laminate, while still maintaining an extreme improvement in the laminate damping properties.

A multiscale method for nonlocal mechanics and diffusion and for the approximation of discontinuous functions

A multiscale implementation of hybrid continuous/discontinuous finite element discretizations of nonlocal models for mechanics and diffusion in two dimensions is developed. The implementation features adaptive mesh refinement based on the detection of defects and results in an abrupt transition between refined elements that contain defects and unrefined elements free of defects. An additional difficulty overcome in the implementation is the design of accurate quadrature rules for stiffness matrix construction that are valid for any combination of the grid size and horizon parameter, the latter being the extent of nonlocal interactions. As a result, the methodology developed can attain optimal accuracy at very modest additional costs relative to situations for which the solution is smooth. Portions of the methodology can also be used for the optimal approximation, by piecewise linear polynomials, of given functions containing discontinuities. Several numerical examples are provided to illustrate the efficacy of the multiscale methodology.