2D Projection Research Articles

Local Attention Transformer-Based Full-View Finger-Vein Identification

Multi-view finger-vein recognition technology has attracted increasing attentions in recent years. Despite recent advances in the multi-view finger-vein identification, existing solutions employ multiple monocular cameras from different views to record two-dimensional (2D) projections of 3D vein vessels, which causes the following problems: 1) 2D images collected from limited views (two or three views) are insufficient for robust 3D vein vessel feature representation. Furthermore, image sequences of the same finger acquired from different views usually show significant differences. As a result, the existing works are still sensitive to positional variations of the fingers, specifically those caused by finger roll movements. 2) Using multiple cameras can lead to increased costs. Moreover, it is impossible to employ several cameras to acquire full-view images because of the limited space on capturing devices. To address the above issues, we present FV-LT, a Full-View Finger-Vein identification system based on a Local attention Transformer, by implementing an image acquisition device with a single camera. First, we design and implement a finger-vein acquisition prototype device that utilizes a single camera and a LED group to rotate along a finger for full-view image collection. This allows capturing all vein patterns concealed beneath human skin to form a complete representation of finger features. Second, given the full-view vein images, we propose a local attention transformer-based approach to extract dependency features of a token (a patch or an image) on its neighborhood’s tokens among image patches and among full-view images, respectively. These dependency features are shown to be robust to positional variations induced by finger rolls. Based on the public database of full-view finger-vein images captured by our designed device, we verify the performance of the proposed FV-LT. The experimental results show that FV-LT significantly outperforms existing 2D/multi-view based approaches with respect to improving the tolerance against finger roll and achieving the state-of-the-art identification accuracy.

Exploratory Analysis of the Gene Expression Matrix Based on Dual Conditional Dimensionality Reduction.

One of the major goals in gene expression data analysis is to explore and discover groups of genes and groups of biological conditions with meaningful relationships. While this problem can be addressed by algorithms, their results require an analysis within context, since they may be affected by many side processes -such as tissue differentiation- that could hinder the target goal. Visual analytics-based methods for exploratory analysis of the gene expression matrix (GEM) are essential in biomedical research since they allow us to frame the analysis within the user's knowledge domain. In this paper, we present a visual analytics approach to discover relevant connections between genes and samples based on linking a reordered GEM heatmap and dual 2D projections of its rows and columns, which can be recomputed conditioned by subsets of genes and/or samples selected by the user during the analysis. We demonstrate the capability of our approach to discover relevant knowledge in three case studies involving two cancer types plus normal tissue from the TCGA database.

Three-dimensional contrast-transfer-function approach in phase-contrast tomography.

A new method is developed for 3D reconstruction of multimaterial objects using propagation-based x-ray phase-contrast tomography (PB-CT) with phaseretrieval via contrast-transfer-function (CTF) formalism. The approach differs from conventional PB-CT algorithms, which apply phaseretrieval to individual 2D projections. Instead, this method involves performing phaseretrieval to the CT-reconstructed volume in 3D. The CTF formalism is further extended to the cases of partially coherent illumination and strongly absorbing samples. Simulated results demonstrate that the proposed post-reconstruction CTF method provides fast and stable phaseretrieval, producing results equivalent to conventional pre-reconstruction 2D CTF phaseretrieval. Moreover, it is shown that application can be highly localized to isolated objects of interest, without a significant loss of quality, thus leading to increased computational efficiency. Combined with the extended validity of the CTF to greater propagation distances, this method provides additional advantages over approaches based on the transport-of-intensity equation.

P‐29: Noise reduction and defect detection algorithm in white‐light scanning interferometry for display panel inspection

White‐light scanning interferometry (WSI) has been widely used for the inspection of various display panels by 3D surface imaging from the micrometer scale to the nanometer scale. However, WSI often fails to sense light reflected from towering defects due to diffused reflection, represented by near‐zero heights, called black spot (b‐spot). It is challenging to detect the b‐spot defect especially in display pad area with varying heights. In addition, vibration of WSI facility hinders a defect spot due to harsh noise amplification. To resolve the issues, we proposed a novel and efficient image processing algorithm which consists of five steps. The first step is image denoising by median filtering with rectangle kernels. The second is pad area preprocessing to avoid misdetection. The third is to identify the presence of b‐spot defect using curvature of 1D projection array of an image. The fourth is defect detection through minimum value finding method. The last is Laplace fitting to infer the height of towering b‐spot defect. We conducted an experimental study using WSI raw data of defective display panels to evaluate the defect detection rate and the reliability of inferred height. We also calculated the correlation between the inferred height by Laplace fitting and measured height by atomic force microscopy (AFM) as a gold standard. We have shown that the proposed method can effectively detect a towering b‐spot defect position with much improved detection rate and infer the height with high precision.

EXAMPLES OF INNOVATIVE METHODS OF TEACHING COMPUTER GRAPHICS SUBJECTS (RESULTS AND DISCUSSION)

This article presents examples, results, and discussion of innovative methods for teaching computer graphics topics. The article examines teaching methods in the field of computer graphics and offers reflections on their innovative approaches. The article begins with the importance of virtual developments in the teaching of computer graphics subjects. Virtual technologies, such as Virtual Reality (VR) and Augmented Reality (AR), allow providing interactive and immersive ways to study computer graphics. Examples show results and discussions of VR in teaching for different fields. In the next part, the article reviews the available techniques of interactive teaching. For example, in interactive graphic design classes, students participate in the creation process of computer graphics programs and work with them to solve basic problems in programming and creation. This method helps you get results according to the schedule and guided counseling during the student creation process. In the next part of the article, innovative aspects of collaborative learning are presented. Examples show how to develop computer graphics projects collaboratively and ensure better learning outcomes through pooled resources. In doing so, students can use them to manage their work, enjoy it, and evaluate their work. In the last part of the article, attention is paid to the presentation of innovative methods of distance education in the field of computer graphics. Distance learning enables students to learn through video conferencing, webinars, and online platforms. This method is a modern way of learning computer graphics and allows students to learn with a constant volunteer.

Retrieving mean volumetric properties of multiphase flows from 2D images: A new approach combining deep learning algorithms and 3D modelling

Measuring the morphological properties of complex multiphase systems is a crucial problem in many areas of science and industry and is particularly difficult in dense environments with limited optical access. This paper presents a new approach capable of extracting three-dimensional (3D) information from spherical particle systems based solely on two-dimensional (2D) projections of the system. Synthetic images of the system are generated using a stochastic geometrical model from a simulated 3D particle system with the same geometrical features as the studied system, which is projected into 2D images labeled with the appropriate 3D information. These images are then fed to a convolutional neural network (CNN) for training before being tested on synthetic and experimental images. Validation results show that this technique successfully predicts the mean features of the studied systems, even for dense environments with overlapping particles, with high computational efficiency.

The optimal triangulation method is not really optimal

AbstractTriangulation refers to the problem of finding a 3D point from its 2D projections on multiple images taken at different camera poses. For solving this problem, it is a common practice to use the so‐called optimal triangulation method. But, the method can be optimal only if we assume no uncertainty in the camera parameters. In this paper, an extensive comparison between various existing methods for the calibrated cameras is performed, in which different poses are considered. Furthermore, uncertainty sensitivity analysis is conducted for extrinsic parameters of cameras. The results show that the optimal triangulation method is actually not the best choice when there is uncertainty in extrinsic parameters. Interestingly, it can be observed that the simple midpoint method works equally well or outperforms the optimal triangulation and many other methods. Apart from its high performance, the midpoint method has a simple closed form solution for multiple views while the optimal triangulation method is hard to be used for more than two views. Therefore, in contrast to the common practice, we argue that the simple midpoint method can be a good choice in the structure‐from‐motion process where there is uncertainty in extrinsic camera parameters.

High-accuracy pose estimation algorithm with monocular camera based on residual network

Pose estimation is a typical problem encountered in computer vision. Therefore, it is important to improve the accuracy of this method. Aiming at the accuracy of pose estimation, we propose a high-accuracy pose estimation algorithm based on the you only look once network and residual network with a monocular camera as the sensor for visual acquisition. This algorithm uses ArUco markers as a reference for object localization and uses the red, green, and blue (RGB) image as the input. The input image is sampled 16 and 32 times to extract the feature image, and the feature image extracted by 16 times sampling is passed through the pass-through layer and then combined with the feature image extracted by 32 times sampling to accomplish the dimension expansion. The feature image is identified by the convolutional layer. The EPnP algorithm is used to solve the camera poses. The pose information of the target object in the RGB image is used as the output. By comparing the pose estimation accuracy for the LINEMOD dataset with three evaluation metrics—the 2D projection metric, ADD metric, and 5 cm to 5 deg metric—it can be observed that the pose estimation algorithm proposed has advantages in terms of accuracy compared with traditional pose estimation algorithms. When the target is very similar to the background objects, the algorithm also achieves good performance.

Extension of King’s Iterative Scheme by Means of Memory for Nonlinear Equations

We developed a new family of optimal eighth-order derivative-free iterative methods for finding simple roots of nonlinear equations based on King’s scheme and Lagrange interpolation. By incorporating four self-accelerating parameters and a weight function in a single variable, we extend the proposed family to an efficient iterative scheme with memory. Without performing additional functional evaluations, the order of convergence is boosted from 8 to 15.51560, and the efficiency index is raised from 1.6817 to 1.9847. To compare the performance of the proposed and existing schemes, some real-world problems are selected, such as the eigenvalue problem, continuous stirred-tank reactor problem, and energy distribution for Planck’s radiation. The stability and regions of convergence of the proposed iterative schemes are investigated through graphical tools, such as 2D symmetric basins of attractions for the case of memory-based schemes and 3D stereographic projections in the case of schemes without memory. The stability analysis demonstrates that our newly developed schemes have wider symmetric regions of convergence than the existing schemes in their respective domains.

RICT: Rotating image computed tomography with a one-to-one reversible image rotation algorithm.

The Mueller, Siddon and Joseph weighting algorithms are frequently used for projection and back-projection, which are relatively complicated when they are implemented in computer code. This study aims to reduce the actual complexity of the projection and back-projection. First, we neglect the exact shape of the pixel, so that its shadow is a rectangle projecting precisely to a detector bin, which implies that all the pixel weights are exactly 1 for each ray through them, otherwise are exactly 0. Next, a one-to-one reversible image rotation algorithm (RIRA) is proposed to compute the projection and back-projection, where two one-to-one mapping lists namely, U and V, are used to store the coordinates of a rotated pixel and its corresponding new coordinates, respectively. For each 2D projection, the projection is simply the column sum in each orientation according to the lists U and V. For each 2D back-projection, it is simply to arrange the projection to the corresponding column element according to the lists U and V. Thus, there is no need for an interpolation in the projection and back-projection. Last, a rotating image computed tomography (RICT) based on RIRA is proposed to reconstruct the image. Experiments show the RICT reconstructs a good image that is close to the result of filtered back-projection (FBP) method according to the RMSE, PSNR and MSSIM values. What's more, our weight, projection and back-projection are much easier to be implemented in computer code than the FBP method. This study demonstrates that the RIRA method has potential to be used to simplify many computed tomography image reconstruction algorithms.

Adapted in place: A digital geometric analysis of the muqarnas dome of al-Bimaristan al-Nuri in Damascus

The dome of al-Bimaristan al-Nuri is regarded as the first muqarnas dome in Damascus. It is a unique example of a suspended muqarnas dome in the region. The 2D projection plane of the dome displays non-uniform cells, which has led researchers to assume that it was an experimental attempt to replicate a new architectural element. This study aims to deliver an in-depth 3D analysis of construction irregularities and deviations to examine the experimental nature of the design and execution. It provides a 3D workflow to document, analyze, and visualize a muqarnas dome. By exploring the geometrical properties and relations of the dome using terrestrial laser scanning (TLS), the study presents geometrical corroboration of the in-situ adaptation of the dome.

Runge–Lenz vector as a 3d projection of SO(4) moment map in phase space

We show, using the methods of geometric algebra, that Runge–Lenz vector in the Kepler problem is a 3-dimensional projection of SO(4) moment map that acts on the phase space of 4-dimensional particle motion. Thus, Runge–Lenz vector is a consequence of geometric symmetry of phase space.

Three-dimensional structure of hybrid magnetic skyrmions determined by neutron scattering

Magnetic skyrmions are topologically protected chiral spin textures which present opportunities for next-generation magnetic data storage and logic information technologies. The topology of these structures originates in the geometric configuration of the magnetic spins - more generally described as the structure. While the skyrmion structure is most often depicted using a 2D projection of the three-dimensional structure, recent works have emphasized the role of all three dimensions in determining the topology and their response to external stimuli. In this work, grazing-incidence small-angle neutron scattering and polarized neutron reflectometry are used to determine the three-dimensional structure of hybrid skyrmions. The structure of the hybrid skyrmions, which includes a combination of N\'eel-like and Bloch-like components along their length, is expected to significantly contribute to their notable stability, which includes ambient conditions. To interpret the neutron scattering data, micromagnetic simulations of the hybrid skyrmions were performed, and the corresponding diffraction patterns were determined using a Born approximation transformation. The converged magnetic profile reveals the magnetic structure along with the skyrmion depth profile, including the thickness of the Bloch and N\'eel segments and the diameter of the core.

SproutAngio: an open-source bioimage informatics tool for quantitative analysis of sprouting angiogenesis and lumen space

Three-dimensional image analyses are required to improve the understanding of the regulation of blood vessel formation and heterogeneity. Currently, quantitation of 3D endothelial structures or vessel branches is often based on 2D projections of the images losing their volumetric information. Here, we developed SproutAngio, a Python-based open-source tool, for fully automated 3D segmentation and analysis of endothelial lumen space and sprout morphology. To test the SproutAngio, we produced a publicly available in vitro fibrin bead assay dataset with a gradually increasing VEGF-A concentration (https://doi.org/10.5281/zenodo.7240927). We demonstrate that our automated segmentation and sprout morphology analysis, including sprout number, length, and nuclei number, outperform the widely used ImageJ plugin. We also show that SproutAngio allows a more detailed and automated analysis of the mouse retinal vasculature in comparison to the commonly used radial expansion measurement. In addition, we provide two novel methods for automated analysis of endothelial lumen space: (1) width measurement from tip, stalk and root segments of the sprouts and (2) paired nuclei distance analysis. We show that these automated methods provided important additional information on the endothelial cell organization in the sprouts. The pipelines and source code of SproutAngio are publicly available (https://doi.org/10.5281/zenodo.7381732).

Observed dust surface density across cosmic times

ABSTRACT Our ability to interpret observations of galaxies and trace their stellar, gas, and dust content over cosmic time critically relies on our understanding of how the dust abundance and properties vary with environment. Here, we compute the dust surface density across cosmic times to put novel constraints on simulations of the build-up of dust. We provide observational estimates of the dust surface density consistently measured through depletion methods across a wide range of environments, going from the Milky Way up to z = 5.5 galaxies. These conservative measurements provide complementary estimates to extinction-based observations. In addition, we introduce the dust surface density distribution function – in analogy with the cold gas column density distribution functions. We fit a power law of the form log f(ΣDust) = −1.92 × log ΣDust − 3.65, which proves slightly steeper than that for neutral gas and metal absorbers. This observed relation, which can be computed by simulations predicting resolved dust mass functions through 2D projection, provides new constraints on modern dust models.

Cometary dust collected by MIDAS on board Rosetta

Context. The Micro-Imaging Dust Analysis System (MIDAS) atomic force microscope (AFM) on board the Rosetta comet orbiter has been dedicated to the collection and 3D topographical investigation of cometary dust in the size range of a few hundreds of nanometers to tens of micrometers with a resolution down to a few nanometers. Aims. We aim to catalog all dust particles collected and analyzed by MIDAS, together with their main statistical properties such as size, height, basic shape descriptors, and collection time. Furthermore, we aim to present the scientific results that can be extracted from the catalog, such as the size distribution and statistical characteristics of cometary dust particles. Methods. Through a careful re-analysis of MIDAS AFM images, we make a significant update and improvement to the existing MIDAS particle catalog, resulting in the addition of more particles and newly developed shape descriptors. The final product is a comprehensive list of all possible cometary dust particles detected by MIDAS. The catalog documents all images of identified dust particles and includes a variety of derived information tabulated one record per particle. Furthermore, the best image of each particle was chosen for subsequent studies. Finally, we created dust coverage maps and clustering maps of the MIDAS collection targets and traced any possible fragmentation of collected particles with a detailed algorithm. Results. The revised MIDAS catalog includes 3523 MIDAS particles in total, where 1857 particles are expected to be usable for further analysis (418 scans of particles before perihelion + 1439 scans of particles after perihelion, both after the removal of duplicates), ranging from about 40 nm to about 8 μm in size. The mean value of the equivalent radius derived from the 2D projection of the particles is 0.91 ± 0.79 μm. A slightly improved equivalent radius based on the particle’s volume coincides in the range of uncertainties with a value of 0.56 ± 0.45 μm. We note that those sizes and all following MIDAS particle size distributions are expected to be influenced by the fragmentation of MIDAS particles upon impact on the collection targets. Furthermore, fitting the slope of the MIDAS particle size distribution with a power law of a · rb yields an index b of ~−1.67 to −1.88. Lastly, based on the created dust coverage maps and clustering maps of the MIDAS collection targets, we determined the particle fragmentation ratio of 4.09 for nominal activity and 11.8 for the outburst, which underlines that parent particles with faster impact velocity are more likely to be fragmented during dust collection.

Deep feature annotation by iterative meta-pseudo-labeling on 2D projections

The absence of large annotated datasets to train deep neural networks (DNNs) is an issue since manual annotation is time-consuming, expensive, and error-prone. Semi-supervised learning techniques can address the problem propagating pseudo labels from supervised to unsupervised samples. However, they still require training and validation sets with many supervised samples. This work proposes a methodology, namely Deep Feature Annotation (DeepFA), that dismisses the validation set and uses very few supervised samples (e.g., 1% of the dataset). DeepFA modifies the feature spaces of a DNN along with meta-pseudo-labeling iterations in a 2D non-linear projection space using the most confidently labeled samples of an optimum-path forest semi-supervised classifier. We present a comprehensive study on DeepFA and a new variant that detects the best DNN model for generalization during the pseudo-labeling iterations. We evaluate components of DeepFA on eight datasets, finding the best DeepFA approach and showing that it outperforms self-pseudo-labeling.

Compressive properties of parametrically optimised mechanical metamaterials based on 3D projections of 4D geometries

The design process of 3D mechanical metamaterials is still an emerging field and in this paper, we propose for the first time, a new design and optimisation approach based on 3D projections of 4D geometries (4-polytopes) and evolutionary algorithms. We find that through iterative parametric optimisation, 4-polytope projected mechanical metamaterials can be optimised to achieve both high specific stiffness and high specific yield strengths. Samples manufactured using a low-stereolithography method were tested in compression. We find that optimised tesseracts (8-cell structures) had a higher specific yield strength (22.8 kN m/kg) than that of honeycomb structures tested out-of-plane (19.4 kN m/kg) and a specific stiffness of (0.68 MN m/kg) which is more than 3-fold that of gyroid structures. The compressive strength to solid-modulus ratio of the 8-cell tesseract is very high (3×10−3), exceeding that of out-of-plane honeycombs, which are themselves closer in value to 5-cell pentatopes (2×10−3). 8-cell and 5-cell structures are in the region of one order of magnitude higher than 16-cell and 24-cell structures (∼2×10−4–8×10−4) and are hence comparable to nanostructured metamaterials. The 8-cell tesseracts are 18% stiffer, 43% stronger, and 19% tougher in compression than out-of-plane honeycomb structures, but unlike honeycombs, 8-cell tesseracts are 3D structures with cubic symmetry. Architecture has a profound effect on the relative consistency of properties with cubically symmetric structures displaying the greatest levels of consistency in terms of both strength and stiffness reduction as a function of pore space. The results presented in this paper showcase the potential of this new class of mechanical metamaterial based on 3D projected 4-polytopes.

Adaptive autoencoder latent space tuning for more robust machine learning beyond the training set for six-dimensional phase space diagnostics of a time-varying ultrafast electron-diffraction compact accelerator.

We present a general adaptive latent space tuning approach for improving the robustness of machine learning tools with respect to time variation and distribution shift. We demonstrate our approach by developing an encoder-decoder convolutional neural network-based virtual 6D phase space diagnostic of charged particle beams in the HiRES ultrafast electron diffraction (UED) compact particle accelerator with uncertainty quantification. Our method utilizes model-independent adaptive feedback to tune a low-dimensional 2D latent space representation of ∼1 million dimensional objects which are the 15 unique 2D projections (x,y),...,(z,p_{z}) of the 6D phase space (x,y,z,p_{x},p_{y},p_{z}) of the charged particle beams. We demonstrate our method with numerical studies of short electron bunches utilizing experimentally measured UED input beam distributions.

GRAPE: A Stochastic Geometrical 3D Model for Aggregates of Particles With Tunable 2D Morphological Projected Properties

The main goal of this paper is to propose a method for the 3D morphological characterization of compact aggregates using 2D image analysis. The problem at hand is the 3D morphometric characterization of latex nanoparticle aggregates. The only available information is 2D projection images of these aggregates, one projection per aggregate. In this context, a method to estimate the 3D morphological characteristics of an aggregate such as the Volume, Surface Area or Solidity from a single projection is proposed. This method is based on a stochastic geometric model called GRAPE (Geometrical Random Aggregation of Particles Emulation) and requires some strong assumptions, and in particular prior estimation of the volume. The model is based on an iterative packing of spheres of identical radii. For each iteration, a fitting function allows to reach objectives corresponding to the desired 2D properties (Area, Perimeter, Aspect Ratio, ...). In order to implement the method, an optimization process must be performed on two parameters of the model: the radius of the elementary particles r and an overlapping distance di. As a validation, this process will be applied to synthetic aggregates, themselves generated from the GRAPE model, then to a population of 104 synthetic aggregates, and finally to 3D printed aggregates whose 3D morphological properties are known thanks to an STL file, and whose projected images have been produced using a morphogranulometer. The results obtained show an excellent approximation of 2D properties by the GRAPE model, and very good results for 3D properties, with less than 5% error on average and less than 2% error in most cases.