Adaptive Multiscale Method Research Articles

Development of an optimal adaptive finite element stabiliser for the simulation of complex flows

An optimal adaptive multiscale finite element method(AMsFEM) for numerical solutions of flow problems modelled by the Oldroyd B model is developed. Complex flows experience instabilities due to a phenomena known as the high Weissenberg number problem. The stabilisers are terms in-cooperated into the variational formulation when applying the finite element method. For the selected few stabilisers, numerical experiments are performed to study the convergence of the solutions. These demonstrate that adaptive strategies reduce the computational load of flow simulation. A best performing combination of choice of stabiliser and adaptive strategy is suggested.

A Dual-Branch Fusion of a Graph Convolutional Network and a Convolutional Neural Network for Hyperspectral Image Classification.

Semi-supervised graph convolutional networks (SSGCNs) have been proven to be effective in hyperspectral image classification (HSIC). However, limited training data and spectral uncertainty restrict the classification performance, and the computational demands of a graph convolution network (GCN) present challenges for real-time applications. To overcome these issues, a dual-branch fusion of a GCN and convolutional neural network (DFGCN) is proposed for HSIC tasks. The GCN branch uses an adaptive multi-scale superpixel segmentation method to build fusion adjacency matrices at various scales, which improves the graph convolution efficiency and node representations. Additionally, a spectral feature enhancement module (SFEM) enhances the transmission of crucial channel information between the two graph convolutions. Meanwhile, the CNN branch uses a convolutional network with an attention mechanism to focus on detailed features of local areas. By combining the multi-scale superpixel features from the GCN branch and the local pixel features from the CNN branch, this method leverages complementary features to fully learn rich spatial-spectral information. Our experimental results demonstrate that the proposed method outperforms existing advanced approaches in terms of classification efficiency and accuracy across three benchmark data sets.

AMSLS: Adaptive multi-scale level set method based on local entropy for image segmentation

Intensity inhomogeneity often appears in medical images and causes great difficulties in image segmentation. Most active contour models perform poorly when applied to intensity inhomogeneous images because their energy functions use local intensity information in a fixed-size domain, causing the contour to evolve in the wrong direction. To overcome the difficulties caused by intensity inhomogeneity, we propose an adaptive multi-scale Gaussian kernel function based on image local entropy, which can determine the appropriate scale for each local region. Choosing the small scale and large scale for inhomogeneous and homogeneous areas respectively make the contour move toward the target boundary accurately. We also propose three adaptive multi-scale (AMS) models, AMS-region scalable fitting (AMS-RSF) model, AMS-local image fitting (AMS-LIF) model, AMS-local and global intensity fitting (AMS-LGIF) model, to segment medical images with intensity inhomogeneity and noise, including left atrial MR images and breast ultrasound images. The experimental results show that the adaptive multi-scale Gaussian kernel function enables the active contour model to effectively segment intensity inhomogeneous images and has a certain robustness to the initial contour and noise, which achieves good performance on MR left atrial images and ultrasound images of breast cancer. The AMS-LGIF model obtained the highest DICE coefficient of 0.9532, which was better than the 0.9429 obtained by the second-ranked LGIF model to segment left atrial MR images. For segmenting breast ultrasound images, the DICE coefficient is increased by 16% than that of the U-Net++ model.

An alternative 2D ICEEMDAN-based denoising method and its application in processing magnetic anomaly data

Due to environmental noise and human factors, magnetic data collected in the field often contain various noises and interferences that significantly affect the subsequent data processing and interpretation. Empirical Mode Decomposition (EMD), an adaptive multiscale analysis method for nonlinear and non-stationary signals, is widely used in geophysical and geodetic data processing. Compared with traditional EMD, Improved Complete Ensemble EMD with Adaptive Noise (ICEEMDAN) is more effective in addressing the problem of mode mixing. Based on the principles of 1D ICEEMDAN, this paper presents an alternative algorithm for 2D ICEEMDAN, extending its application to two-dimensional scenarios. The effectiveness of the proposed approach is demonstrated through synthetic signal experiments, which show that the 2D ICEEMDAN exhibits a weaker mode mixing effect compared to the traditional bidimensional EMD (BEMD) method. Furthermore, to improve the performance of the denoising method based on 2D ICEEMDAN and preserve useful signals in high-frequency components, an improved soft thresholding technique is introduced. Synthetic magnetic anomaly data testing indicates that our denoising method effectively preserves signal continuity and outperforms traditional soft thresholding methods. To validate the practical application of this improved threshold denoising method based on 2D ICEEMDAN, it is applied to ground magnetic survey data in the Yandun area of Xinjiang. The results demonstrate the effectiveness of the method in removing noise while retaining essential information from practical magnetic anomaly data. In particular, practical applications suggest that 2D ICEEMDAN can extract trend signals more accurately than the BEMD. In conclusion, as a potential tool for multi-scale decomposition, the 2D ICEEMDAN is versatile in processing and analyzing 2D geophysical and geodetic data.

Volume enlargement of the choroid plexus and brain ventricles in drug-naïve, first-episode major depressive disorder

BackgroundWe investigated volumetric alterations in the bilateral choroid plexus (ChP) and brain ventricles of patients during their first episode of major depressive disorder (MDD) prior to antidepressant treatment. MethodsSeventy-one first-episode drug-naïve patients with MDD and seventy-four healthy control (HC) subjects were recruited. MRI data were obtained, and bilateral ChP and brain ventricle volumes were evaluated using segmentation, based on the adaptive multiscale and expectation maximization method. One-way multivariate analysis of covariance was used to calculate volumetric differences in the bilateral ChP and brain ventricles between the groups, and partial Pearson correlation analyses were used to investigate the relationship between the volumes of the bilateral ChP and brain ventricles. ResultsFirst-episode drug-naïve patients with MDD showed enlarged volumes of the bilateral ChP, bilateral lateral ventricle (LV), and third ventricle compared with HCs. The ChP volume positively correlated with the LV and third ventricle, but not with the fourth ventricle in patients with MDD, whereas it correlated with all four brain ventricles in HCs. We did not observe significant correlations between bilateral ChP volume and brain ventricles, HAMD scores, or symptom severity. LimitationsOur study populations differed in age and sex and we did not extensively measure the amount of neuroinflammation in the brain or blood, include a functional assessment, nor evaluate other neural comorbidities or neuropsychiatric conditions. ConclusionsOur study extends the existing research to suggest that illness-related alterations in ChP volume enlargement in first-episode antidepressant-naïve patients with MDD may serve as a trait measure.

Object Detection in Remote Sensing Images Based on Adaptive Multi-Scale Feature Fusion Method

Multi-scale object detection is critical for analyzing remote sensing images. Traditional feature pyramid networks, which are aimed at accommodating objects of varying sizes through multi-level feature extraction, face significant challenges due to the diverse scale variations present in remote sensing images. This situation often forces single-level features to span a broad spectrum of object sizes, complicating accurate localization and classification. To tackle these challenges, this paper proposes an innovative algorithm that incorporates an adaptive multi-scale feature enhancement and fusion module (ASEM), which enhances remote sensing image object detection through sophisticated multi-scale feature fusion. Our method begins by employing a feature pyramid to gather coarse multi-scale features. Subsequently, it integrates a fine-grained feature extraction module at each level, utilizing atrous convolutions with varied dilation rates to refine multi-scale features, which markedly improves the information capture from widely varied object scales. Furthermore, an adaptive enhancement module is applied to the features of each level by employing an attention mechanism for feature fusion. This strategy concentrates on the features of critical scale, which significantly enhance the effectiveness of capturing essential feature information. Compared with the baseline method, namely, Rotated FasterRCNN, our method achieved an mAP of 74.21% ( 0.81%) on the DOTA-v1.0 dataset and an mAP of 84.90% (+9.2%) on the HRSC2016 dataset. These results validated the effectiveness and practicality of our method and demonstrated its significant application value in multi-scale remote sensing object detection tasks.

Fast compressed sensing spectral measurement with adaptive gradient multiscale resolution

We propose a fast, adaptive multiscale resolution spectral measurement method based on compressed sensing. The method can apply variable measurement resolution over the entire spectral range to reduce the measurement time by over 75% compared to a global high-resolution measurement. Mimicking the characteristics of the human retina system, the resolution distribution follows the principle of gradually decreasing. The system allows the spectral peaks of interest to be captured dynamically or to be specified a priori by a user. The system was tested by measuring single and dual spectral peaks, and the results of spectral peaks are consistent with those of global high-resolution measurements.

Adaptive Multi-Scale Fusion Blind Deblurred Generative Adversarial Network Method for Sharpening Image Data

Drone and aerial remote sensing images are widely used, but their imaging environment is complex and prone to image blurring. Existing CNN deblurring algorithms usually use multi-scale fusion to extract features in order to make full use of aerial remote sensing blurred image information, but images with different degrees of blurring use the same weights, leading to increasing errors in the feature fusion process layer by layer. Based on the physical properties of image blurring, this paper proposes an adaptive multi-scale fusion blind deblurred generative adversarial network (AMD-GAN), which innovatively applies the degree of image blurring to guide the adjustment of the weights of multi-scale fusion, effectively suppressing the errors in the multi-scale fusion process and enhancing the interpretability of the feature layer. The research work in this paper reveals the necessity and effectiveness of a priori information on image blurring levels in image deblurring tasks. By studying and exploring the image blurring levels, the network model focuses more on the basic physical features of image blurring. Meanwhile, this paper proposes an image blurring degree description model, which can effectively represent the blurring degree of aerial remote sensing images. The comparison experiments show that the algorithm in this paper can effectively recover images with different degrees of blur, obtain high-quality images with clear texture details, outperform the comparison algorithm in both qualitative and quantitative evaluation, and can effectively improve the object detection performance of blurred aerial remote sensing images. Moreover, the average PSNR of this paper’s algorithm tested on the publicly available dataset RealBlur-R reached 41.02 dB, surpassing the latest SOTA algorithm.

An Adaptive Multiscale Gaussian Co-Occurrence Filtering Decomposition Method for Multispectral and SAR Image Fusion

An Adaptive Multiscale Gaussian Co-Occurrence Filtering Decomposition Method for Multispectral and SAR Image Fusion

Performance analysis of different DCNN models in remote sensing image object detection

In recent years, deep learning, especially deep convolutional neural networks (DCNN), has made great progress. Many researchers use different DCNN models to detect remote sensing targets. Different DCNN models have different advantages and disadvantages. In this paper, we use YoloV4 as the detector to “fine-tune” various mainstream deep convolutional neural networks on two large public remote sensing data sets−LEVIR data set and DOTA data set to compare the advantages of various networks. This paper analyzes the reasons why the effect of “fine-tuning” convolutional neural networks is sometimes not good, and points out the difficulties of object detection in optical remote sensing images. To improve the detection accuracy of optical remote sensing targets, in addition to “fine-tuning” convolutional neural network, we also provide a variety of adaptive multi-scale feature fusion methods to improve the detection accuracy. In addition, for the large number of parameters generated by deep convolutional neural network, we provide a method to save storage space.

Adaptive multi-scale feature fusion based U-net for fracture segmentation in coal rock images

Accurate segmentation of fractures in coal rock CT images is important for the development of coalbed methane. However, due to the large variation of fracture scale and the similarity of gray values between weak fractures and the surrounding matrix, it remains a challenging task. And there is no published dataset of coal rock, which make the task even harder. In this paper, a novel adaptive multi-scale feature fusion method based on U-net (AMSFF-U-net) is proposed for fracture segmentation in coal rock CT images. Specifically, encoder and decoder path consist of residual blocks (ReBlock), respectively. The attention skip concatenation (ASC) module is proposed to capture more representative and distinguishing features by combining the high-level and low-level features of adjacent layers. The adaptive multi-scale feature fusion (AMSFF) module is presented to adaptively fuse different scale feature maps of encoder path; it can effectively capture rich multi-scale features. In response to the lack of coal rock fractures training data, we applied a set of comprehensive data augmentation operations to increase the diversity of training samples. These extensive experiments are conducted via seven state-of-the-art methods (i.e., FCEM, U-net, Res-Unet, Unet++, MSN-Net, WRAU-Net and ours). The experiment results demonstrate that the proposed AMSFF-U-net can achieve better segmentation performance in our works, particularly for weak fractures and tiny scale fractures.

An adaptive multiscale finite element method for strain localization analysis with the Cosserat continuum theory

Strain localization analysis based on finite element method (FEM) usually requires an intensive computation to capture an accurate shear band and the limit stress, especially in the heterogeneous problems, and thus costs much computational resource and time. This paper proposes an adaptive multiscale FEM (AMsFEM) to improve the computational efficiency of strain localization analysis in heterogeneous solids. In the multiscale analysis, h- and p-adaptive strategies are proposed to update the fine and coarse meshes, respectively. The problem of mesh dependence is handled by the Cosserat continuum theory. In the fine-scale adaptive procedure, triangular elements are taken to discretize the fine-scale domain, and the newest vertex bisection is utilized for refinement based on the gradient of displacement. In the coarse-scale adaptive procedure, a multi-node coarse element technique is considered. By introducing a probability density function for each side of a coarse element, the optimal positions of the newly added coarse nodes can be determined. With the proposed adaptive multiscale procedure, the computational DOFs are reduced smartly and massively. Three representative heterogeneous examples demonstrate that the proposed method can accurately capture shear bands, with an improved computational efficiency and robust convergence.

Adaptive multi-scale beam lattice method for competitive trans-scale crack growth simulation of heterogeneous concrete-like materials

In order to overcome the shortcomings of traditional pure macro- and meso-scale crack growth simulation method, here an adaptive multi-scale beam lattice method (AMBLM) is developed to simulate trans-scale crack growth process from meso-scale to macro-scale for quasi-brittle heterogeneous concrete-like materials with low computation cost. In the trans-scale crack growth simulation, dangerous regions can be automatically found and the meso-scale model is then re-modeled in the found region, which can ensure continuous changes in the adaptive multi-scale modeling due to cracking. In this way, the meso-scale simulation is only implemented in the dangerous region of crack growth paths, and macro-scale simulation is implemented in other regions. This can improve computational efficiency of the developed method significantly. In order to verify the developed AMBLM, the trans-scale crack growth process of two concrete specimens are simulated. Then the simulation results are compared with experimental results, which show that the developed AMBLM is effective. In addition, the simulation results are also compared with the computation results obtained from pure meso-scale simulation. The results show that AMBLM has very high precision and superior computational efficiency on multi-scale crack growth simulation of quasi-brittle concrete-like materials.

River body extraction from sentinel-2A/B MSI images based on an adaptive multi-scale region growth method

River networks are important water carriers that provide a multitude of ecosystem services, including freshwater for agriculture, drinking water for cities, and recreational activities. Accurate mapping of river networks from remote-sensing images is important for the study of these systems. Unfortunately, the delineation of river networks is challenging due to the meandering nature of river channels, the often complex and variable features of the surrounding landscape, and the spatial heterogeneity of the river networks. Here, we present an adaptive, multi-scale region growth method (AMRGM) to delineate river networks from sentinel-2A/B MSI images. The method can handle variable river heterogeneous surroundings, multiple spatial scales, and variable curvatures of the river channels. The method includes four steps: (1) a water index (NDWI) is used to provide initial detection of river water pixels in the image; (2) a bias-corrected fuzzy C-means (BCFCM) method alleviates the effects of the variable surrounding landscape; (3) a scale-enhancement algorithm based on the hessian matrix makes full use of scale and direction information to enhance river morphology characteristics (multiple dimensions and variable curvatures), and (4) a regional growth criterion facilitates handling of various river dimensions. Fast-growing and fine-screening strategies are also included in the AMRGM. The method is applied to eight river networks to evaluate its accuracy and reliability with various river morphologies and hydrological conditions. The AMRGM is more widely applicable than four commonly used river-detection methods (i.e., K-means method, maximum likelihood method, iterative self-organizing data analysis technique algorithm, and support vector machine) and outperform these methods when detecting multi-scale river branches. The mean overall accuracy (OA) and kappa coefficients (KC) of the AMRGM exceed 97% and 0.92 across the eight river networks. The most accurate river extractions are obtained for large rivers such as the Amazon River, Mackenzie River, and Ganges River Delta, which have more discernable scale and direction characteristics. Relatively high omission and commission errors are observed in river networks with complex and heterogeneous zonations, such as the river Welland, UK, and the Zagya Zangbo River in the Tibet plateau. The complex geomorphic features of the river Welland reduce OA and KC to 93.8% and 0.86, respectively

High-accuracy calibration of cameras without depth of field and target size limitations.

During camera calibration, targets need to be placed in the depth of field of the lens to ensure clear imaging, and they should take up proper proportions in the image. These requirements cause difficulty in many calibration scenarios, such as those involving large-field-of-view, shallow-depth-of-field, or online operation cameras. In view of the above-mentioned problems, this study proposes a high-accuracy camera calibration method, which can overcome the influence of image blur and noise and is not limited by depth of field and target size. First, a high-accuracy light-spot small target is placed closely in front of the camera, so that the target image can take up a large proportion in the whole image. In case of defocus blur, the adaptive multi-scale method is used to extract feature point coordinates at first to ensure accuracy, and the location variance of each feature point is estimated concurrently. Finally, the high-accuracy intrinsic and extrinsic parameters of the camera under test are obtained by nonlinear optimization where re-projection errors are normalized by location variances according to the Gauss-Markov theorem. Simulation and physical experiments validate the effectiveness of the proposed method.

Infrared dim target detection method inspired by human vision system

Infrared dim target detection has long been a key technology for various systems, such as infrared search and track (IRST) systems and the Space-Based Infrared System (SBIRS). However, it is difficult for traditional detection methods to adapt to different types of complex backgrounds. Therefore, this paper proposes an adaptive infrared dim target detection method based on human visual contrast, motion, prediction, and other characteristics. First, according to the characteristics of different types of background images, the classification preprocessing strategy is adopted to remove noise, suppress the background, and improve the target signal-to-noise ratio. Second, on the basis of the visual contrast and scale adaptation mechanism, we propose an adaptive multi-scale local contrast method to extract the saliency region, and we then analyze the spectral scale to further suppress the background, enhance the target central area, and construct a suspected target set. Finally, the candidate moving target set is obtained by motion region matching using the optical flow method, and a multi-frame screening strategy combined with dynamic pipeline filtering is proposed to identify the target and reduce the false positive rate. Our experiment results indicate that the proposed method can adapt to changes in the target scale and achieve stable and adaptive detection of dim targets in the background of sky, sea-sky, and ground objects.

Image Edge Recognition of Virtual Reality Scene Based on Multi-Operator Dynamic Weight Detection

Image edge detection in virtual reality scenes is one of the most important technologies in the field of image processing and computer vision. It occupies an important position in the image processing system and is a key factor that affects the performance of the entire system. The quality of the algorithm directly affects the performance of the computer's visual system. This paper briefly describes the basic theoretical methods of image edge detection. The basic theory and processing flow of image edge detection are analyzed, and several commonly used edge detection preprocessing methods and their basic principles are elaborated in detail. The existing traditional edge detection methods are briefly introduced, and the traditional edge detection methods are compared. This paper studies the adaptive multi-scale edge detection method based on Canny algorithm, and through theoretical analysis, compares the advantages and disadvantages of various algorithms in image edge detection. In the process of acquiring image edges, the local maximum value of the modulus is calculated along the gradient direction and the threshold is adaptively selected based on the image blocking principle. Comparing with the traditional modulus maximum edge detection method, this algorithm overcomes the contradiction that can suppress the interference of noise and obtain better edge detection effect to a certain extent. Experiments show that the method in this paper not only suppresses the interference of impulsive noise to the image edge detection, but also greatly reduces the possibility of false edges, and obtains a satisfactory accurate binary edge image.

Adaptive numerical homogenization for upscaling single phase flow and transport

We propose an adaptive multiscale method to improve the efficiency and the accuracy of numerical computations by combining numerical homogenization and domain decomposition for modeling flow and transport. Our approach focuses on minimizing the use of fine scale properties associated with advection and diffusion/dispersion. Here a fine scale flow and transport problem is solved in subdomains defined by a transient region where spatial changes in transported species concentrations are large while a coarse scale problem is solved in the remaining subdomains. Away from the transient region, effective macroscopic properties are obtained using local numerical homogenization. An Enhanced Velocity Mixed Finite Element Method (EVMFEM) as a domain decomposition scheme is used to couple these coarse and fine subdomains [1]. Specifically, homogenization is employed here only when coarse and fine scale problems can be decoupled to extract temporal invariants in the form of effective parameters. In this paper, a number of numerical tests are presented for demonstrating the capabilities of this adaptive numerical homogenization approach in upscaling flow and transport in heterogeneous porous medium.

Adaptive generalized multiscale finite element methods for H(curl)-elliptic problems with heterogeneous coefficients

In this paper, we construct an adaptive multiscale method for solving H(curl)-elliptic problems in highly heterogeneous media. Our method is based on the generalized multiscale finite element method. We will first construct a suitable snapshot space, and a dimensional reduction procedure to identify important modes of the solution. We next develop and analyze an a posteriori error indicator, and the corresponding adaptive algorithm. In addition, we will construct a coupled offline–online adaptive algorithm, which provides an adaptive strategy to the selection of offline and online basis functions. Our theory shows that the convergence is robust with respect to the heterogeneities and contrast of the media. We present several numerical results to illustrate the performance of our method.

A three dimensional adaptive multiscale method for crack growth in Silicon

A three dimensional concurrently coupled adaptive multiscale method is introduced here to simulate complex crack growth patterns in Silicon, by combining several numerical techniques across the length scales. The coarse scale material is modeled using the virtual atom cluster approach. The strong kinematic discontinuities in the bulk are simulated based on a three dimensional version of the phantom node method. A molecular statics model placed around the crack tip is concurrently coupled with the phantom-based discontinuous formulation, where the coupling between the fine and coarse scales is realized through the use of ghost atoms, whose positions are interpolated based on the coarse scale solution. The boundary conditions to the fine scale model, at the coupling region, are assigned by enforcing the interpolated displacements of ghost atoms. In order to optimize the computation costs, adaptivity schemes for adjustment of the fine scale region as the crack propagates, and coarse graining of the region behind the crack tip, are proposed. The crack tip location is detected based on an energy criterion. All the molecular simulations in the pure atomistic as well as the multiscale model are carried out using the LAMMPS software, triggered through the system command in MATLAB. The performance of the developed framework in terms of computation cost, robustness and versatility, is assessed through several numerical examples concerning crack growth in Silicon. Therefore, the diamond cubic lattice structure of Silicon is used at the fine scale, where the atom-atom interactions are modeled based on the Tersoff potential function. According to the numerical examples presented in this study, savings in computational time using the present multiscale method are observed to be up to 87%, as compared to the pure atomistic model.