Reconstruction Error Research Articles

Multi-Channel Multi-Scale Convolution Attention Variational Autoencoder (MCA-VAE): An Interpretable Anomaly Detection Algorithm Based on Variational Autoencoder.

With the rapid development of industry, the risks factories face are increasing. Therefore, the anomaly detection algorithms deployed in factories need to have high accuracy, and they need to be able to promptly discover and locate the specific equipment causing the anomaly to restore the regular operation of the abnormal equipment. However, the neural network models currently deployed in factories cannot effectively capture both temporal features within dimensions and relationship features between dimensions; some algorithms that consider both types of features lack interpretability. Therefore, we propose a high-precision, interpretable anomaly detection algorithm based on variational autoencoder (VAE). We use a multi-scale local weight-sharing convolutional neural network structure to fully extract the temporal features within each dimension of the multi-dimensional time series. Then, we model the features from various aspects through multiple attention heads, extracting the relationship features between dimensions. We map the attention output results to the latent space distribution of the VAE and propose an optimization method to improve the reconstruction performance of the VAE, detecting anomalies through reconstruction errors. Regarding anomaly interpretability, we utilize the VAE probability distribution characteristics, decompose the obtained joint probability density into conditional probabilities on each dimension, and calculate the anomaly score, which provides helpful value for technicians. Experimental results show that our algorithm performed best in terms of F1 score and AUC value. The AUC value for anomaly detection is 0.982, and the F1 score is 0.905, which is 4% higher than the best-performing baseline algorithm, Transformer with a Discriminator for Anomaly Detection (TDAD). It also provides accurate anomaly interpretation capability.

An accessible and intuitive approach to 3D data modeling for morphometric surface evaluation in pig farming

The recording of pig surfaces allows for the estimation of growth-related body size dimensions from which the animals’ condition can be inferred. Furthermore, for purposes such as animal breeding, it is of particular interest to describe the shape variations within the population. Therefore, we built a Statistical Shape Model (SSM) that enables a quantitative and visual exploration of variation in size, form and posture in an intuitive way.We included 4315 images from 582animals with a body weight ranging from 50 to 140 kg. After data processing, i.e. the removal of data points corresponding to the head and tail, the 3D point clouds were reduced to an equal number of reference points. These reference points were arranged so that the pig surface could be reconstructed via Bézier surface patches. Then, Procrustes Analysis was performed to align the reference points from all shapes. A mean shape and its main modes of variation were derived, some of which were used to correct the posture of the animals.A linear regression for weight prediction resulted in a mean absolute error of 5.06 kg before and 2.84 kg after data processing. The median error of surface reconstruction was 2.5 mm taking eight surface patches with 45 reference points and 1.3 mm taking 32 surface patches with 153 reference points into account. The first six modes of variation from the SSM indicated differences in size and posture. Interestingly, regions prone to noise were also revealed.In conclusion, we present a novel approach that can be used to receive a compact number of corresponding data points, providing data that is easier and more efficient to handle while exposing shape-specific attributes. Our approach is helpful to visualize and compare shapes, including the collection of body size measurements as well as unsupervised and supervised clustering for more objective data-driven decision-making.

Slow feature‐constrained decomposition autoencoder: Application to process anomaly detection and localization

SummaryDetecting anomalies in manufacturing processes is crucial for ensuring safety. However, noise significantly undermines the reliability of data‐driven anomaly detection models. To address this challenge, we propose a slow feature‐constrained decomposition autoencoder (SFC‐DAE) for anomaly detection in noisy scenarios. Considering that the process can exhibit both long‐term trends and periodic properties, the process data is decomposed into trends and cycles. The repetitive information is mitigated by slicing and randomly masking certain trends and cycles. Dependencies among slices are constructed to extract intrinsic information, while high‐frequency noise is reduced using a slow feature‐constrained loss. Anomalies are detected and localized through a reconstruction error strategy. The effectiveness of SFC‐DAE is demonstrated using data from a sugar factory and a secure water treatment system.

Semi-automated minimization of brick-mortar segmentation errors in 3D historical wall reconstruction

In current practices of geometric modeling and building information modeling (BIM) of architectural heritage, point cloud data acquired from site surveys is segmented into parts that correspond to actual building elements. Deep learning methods are used to automate the segmentation processes; however, each historical building presents unique challenges due to distinct construction techniques, irregular applications, deformations, and wear on the surveyed surfaces, and due to these peculiarities, segmented point cloud data may display local errors in the detection of individual architectural elements. This paper presents a method for post-processing segmented point cloud data to semi-automatically detect building elements such as bricks in historical walls. Results show that this approach, which involves testing neighborhoods of points and classified areas to identify and correct misclassified points, diminishes segmentation errors and reconstructs bricks close to their original form.

Empowering quadrature mirror filter bank architectures: dynamic-grey wolf optimization approach for elevated filter orders

The present study introduces a revised version of the Grey Wolf Optimization (GWO) algorithm called Dynamic-GWO, which integrates a dynamic parameter adaptation mechanism for population diversity enhancement. This mechanism implements the real-time adjustment of control parameters throughout the optimization procedure. The algorithm's effectiveness is demonstrated by its consistent performance across a variety of intricate benchmark functions. Further, the suggested approach is employed to determine the eminent filter tap coefficients of the Quadrature-Mirror-Filter (QMF) bank that enables the configuration with filter orders of 120 and 150, which have yet to be achieved using existing methods. The uniqueness of the proposed approach lies in its capability to overcome the inherent difficulties associated with higher-order filter design, including increased filter length and computational complexity. The primary objective function is to minimize the algebraic sum of three types of distortion: pass-band error, excess power of stop-band, and squared error of the overall transfer function at the quadrature frequency ω q = π / 2 , which generally occur in the procedure of Near-Perfect-Reconstruction (NPR) QMF bank design. In addition, the measured maximum reconstruction errors are 1.303 × 10 − 15 , 2.93 × 10 − 16 and 2.18 × 10 − 06 dB, and aliasing distortions are A L = 85.80 , 128.49 and 153.52 dB for filter orders N = 100, 120 and 150, correspondingly. The proposed algorithm yields significant improvements with an 86.68% reduction in the ripple factor and a 10.45% increase in attenuation compared to the baseline GWO for filter order 120, underscoring its effectiveness in enhancing filter performance. The observed outcomes indicate that the Dynamic-GWO algorithm consistently surpasses the baseline GWO and other Nature-Inspired Algorithms (NIAs) regarding computational ease and convergence speed.

Performance of a reference-free wavefront sensor based on the Talbot effect under Gaussian illumination

A reference-free wavefront reconstruction technique based on the Talbot sensor under Gaussian beam illumination is discussed in the paper. Since pit images are formed in several sequential planes due to the Talbot effect, it allows us to reconstruct the wavefront on the basis of two object self-images in nearby planes without any reference. Simulation results of the reference-free reconstruction of a defocus aberration of optical waves have been compared with the classical Talbot-sensor-based approaches. Additionally, the reconstruction error from the self-image of the gratings and the proposed reconstruction procedure that allows us to decrease wavefront estimation error are analyzed.

High Accuracy Reconstruction of Airborne Streak Tube Imaging LiDAR Using Particle Swarm Optimization

Airborne streak tube imaging LiDAR (STIL) consists of several different data-generating subsystems and introduces system errors each time it is installed on an aircraft. These errors change with each installation, which makes the parametric calibration of the LiDAR meaningless. In this study, we propose a high-precision reconstruction method for point clouds that can be used without calibrating the system parameters. In essence, after each remote sensing measurement, a self-checking process is performed with experimental data to replace the fixed system parameters. In this process, the splicing error of the same region measured under different conditions is used as a criterion to optimize the reconstruction parameters via a particle swarm optimization (PSO) algorithm. For a detection distance of 3000 m, the elevation error of the point cloud reconstruction reaches more than 1 m if the placement parameters are not optimized; after optimization, the elevation error can be controlled within 0.3 m.

A Novel One‐Class Convolutional Autoencoder Combined With Excitation–Emission Matrix Fluorescence Spectroscopy for Authenticity Identification of Food

ABSTRACTIn this work, a novel one‐class classification algorithm one‐class convolutional autoencoder (OC‐CAE) was proposed for the detection of abnormal samples in the excitation–emission matrix (EEM) fluorescence spectra dataset. The OC‐CAE used Boxplot to analyze the reconstruction errors and used the LOF algorithm to handle features extracted by the hidden layer in the convolutional autoencoder (CAE). The fused information provides the basis for more accurate pattern recognition, ensures flexibility in model training, and can obtain higher model specificity, which is important in the field of food quality control. To demonstrate the reliability and advantages of OC‐CAE, two EEM cases related to the authentication of food including the Zhenjiang aromatic vinegar (ZAV) case and the camellia oil (CAO) case were studied. The results showed that OC‐CAE identified all abnormal samples in the two cases, reflecting excellent performance in the detection of abnormal samples, and that it, coupled with EEM, would be an effective tool for the authenticity identification of food.

DDANF: Deep denoising autoencoder normalizing flow for unsupervised multivariate time series anomaly detection

In recent years, the proliferation of IoT technologies and the widespread adoption of wireless sensors across various critical infrastructures such as power plants, service monitoring systems, space and earth exploration missions, and water treatment facilities have resulted in the generation of vast quantities of multivariate time series data. Within this context, unsupervised anomaly detection has emerged as a pivotal yet challenging problem in time series research, necessitating machine learning models capable of identifying rare anomalies amidst massive datasets. Traditionally, unsupervised methods have approached this issue by learning representations of primary patterns within sequences and detecting deviations through reconstruction errors. However, the effectiveness of this approach is often limited due to the intricate dynamics and diverse patterns inherent in these dynamic systems. Moreover, many existing unsupervised anomaly detection techniques fail to fully exploit inter-feature relationships within multivariate time series data, thereby overlooking a crucial criterion for accurate detection. To address these shortcomings, this paper introduces a novel unsupervised method for multivariate time series anomaly detection based on normalized flows and autoencoders. Central to our approach is the incorporation of a channel shuffling mechanism during training, enhancing the model’s capacity to discern inter-channel patterns and anomalies. Concurrently, the application of normalized flows within the autoencoder framework serves to constrain the latent space, effectively isolating anomalies and improving detection accuracy. Experimental validation conducted on two large-scale public datasets demonstrates the efficacy of the proposed method compared to established benchmarks, highlighting its superior performance.

Intelligent skin-removal photoacoustic computed tomography for human based on deep learning.

Photoacoustic computed tomography (PACT) has centimeter-level imaging ability and can be used to detect the human body. However, strong photoacoustic signals from skin cover deep tissue information, hindering the frontal display and analysis of photoacoustic images of deep regions of interest. Therefore, we propose a 2.5 D deep learning model based on feature pyramid structure and single-type skin annotation to extract the skin region, and design a mask generation algorithm to remove skin automatically. PACT imaging experiments on the human periphery blood vessel verified the correctness our proposed skin-removal method. Compared with previous studies, our method exhibits high robustness to the uneven illumination, irregular skin boundary, and reconstruction artifacts in the images, and the reconstruction errors of PACT images decreased by 20% ~ 90% with a 1.65 dB improvement in the signal-to-noise ratio at the same time. This study may provide a promising way for high-definition PACT imaging of deep tissues.

A Standard-Cell-Based Neuro-Inspired Integrate-and-Fire Analog-to-Time Converter for Biological and Low-Frequency Signals - Comparison With Analog Version.

Continuous-time asynchronous data converters namely, analog-to-digital converters and analog-to-time converters, can be beneficial for certain types of applications, such as, processing of biological signals with sparse information. A particular case of these converters is the integrate-and-fire converter (IFC) that is inspired by the neural system. If it is possible to develop a standard-cell-based (SCB) IFC circuit to perform well in advanced technology nodes, it will benefit from the simplicity of SCB circuit designs and can be implemented in widely available field-programmable gate arrays (FPGAs). This way, this paper proposes two IFC circuits designed and prototyped in a 130 nm CMOS standard process. The first is a novel SCB open-loop dynamic IFC. The latter, is a closed-loop analog IFC with conventional blocks. This paper presents a through comparison between the two IFC circuits. They have a power dissipation of 59 μW and 53 μW, and an energy per pulse of 18 pJ and 1060 pJ, SCB and analog IFC, respectively. The SCB IFC has one of the lowest energy per pulse consumption reported for IFC circuits. The analog IFC, being fully differential, is to our knowledge the first of its kind. Moreover, they do not require an external clock. They can convert signals with a peak-to-peak amplitude from 1.6 mV to 28 mV and 0.6 mV to 2.4 mV, and a frequency range of 2 Hz to 42 kHz and 10 Hz to 4 kHz, SCB and analog IFC, respectively. Presenting low normalized RMS conversion plus reconstruction errors, below 5.2%. The maximum pulse density (average firing-rate) is 3300 kHz, for the SCB and 50 kHz, for the analog IFC.

Building Coarse to Fine Convex Hulls With Auxiliary Vertices for Palette-Based Image Recoloring.

Constructing a convex hull for the pixel colors of an image by viewing them as 3D points can extract a set of palette colors for the image, then image recoloring can be achieved by modifying the palette colors. For better recoloring effect, the convex hull should contain more pixels (inclusive) and be more compact. Otherwise, reconstruction error would occur or the extracted palette color would be less representative, yielding wrong recoloring results or less effective edit. We observe that convex hulls constructed by prior methods can contain all the image pixels, but are far from compact. Efforts have been made to optimize the vertices of convex hull to increase the compactness but are still not perfect. In this paper, we propose a novel coarse to fine convex hull construction scheme with auxiliary vertices. We start by constructing a coarse convex hull whose vertices are directly image pixels which is thus the most compact but cannot contain all pixels. We then make a remedy by adding auxiliary vertices into the coarse convex hull to obtain a fine convex hull. More auxiliary vertices are added, more image pixels will be contained into the fine convex hull. The auxiliary vertices are image pixels too so that the compactness can still be maintained. During editing, the auxiliary vertices are not allowed to be edited for edit convenience, but deformed as-rigid-as-possible with the adjusting of other vertices. Our convex hull is both inclusive and compact. Extensive experiments validate the effectiveness of the proposed method.

A global adaptive discretization of velocity space for discrete velocity methods in predictions of rarefied and multi-scale flows

By introducing a discrete velocity space (DVS), deterministic methods in gas-kinetic theory, such as the discrete velocity method (DVM) and unified methods, can accurately capture complex nonequilibrium distribution functions and describe rarefied flow behaviors. However, describing high-speed flows with conventional Cartesian DVS is prohibitively costly due to the large number of discrete velocity points. Therefore, to enable deterministic solvers to handle complex, rarefied, and multi-scale flows effectively, a novel adaptive velocity space (AVS) is proposed. First, a global velocity mesh is intentionally adopted instead of a local velocity mesh to maintain a high level of DVS parallelism and facilitate extension to implicit algorithms. The global AVS is robust as it avoids the instability of information transformation between different cell-local AVS. Second, a new strategy is developed for reconstructing the distribution function in the tree-structured AVS, which is a low-order reconstruction with forced macroscopic conservation. This low-order reconstruction facilitates the direct value assignment between farther and child nodes, avoiding the derivative calculation of the distribution function (which is sometimes discontinuous). Additionally, the integration error of the low-order reconstruction is fixed by the forced macroscopic conservation. With these two important improvements, the proposed global AVS is then integrated into general DVM frameworks, such as the unified gas-kinetic scheme. Finally, a number of numerical tests are carried out to validate the proposed method, including steady and unsteady multi-scale flows.

Comparison of different algorithms to improve the quality of Image

Super-resolution (SR) become the most important task in image processing because of requirement of high quality of images. It becomes popular in technology like computer vision and pattern recognition. In Single Image Super-resolution (SR) uses the sequences of single low resolution (LR) images from high resolution (HR) to improve the quality of image. The SR is an image processing problem which aims to generate a high resolution (HR) image from low resolution input image. The SR is difficult because of the missing information in the given LR image. Thus in SR it is very important to boost the reconstruction performance so that, there is need of effective prior knowledge. The prior knowledge of image will solve the problem to improve the quality of generated image. The main aim of SR is to generate the HR image with good visual perception as similar as original image. The paper compare two methods of SR that is bi-linear method, and direction group sparsity of image method that is proposed one. These methods will evaluate the accuracy, subjective visual effect, quality of image, time complexity and PSNR, etc. These calculations will enhance nature of picture like better visual impact quality, lower reconstruction error and adequate calculation proficiency, and so on

Pose-Driven Compression for Dynamic 3D Human via Human Prior Models.

To cost-effectively transmit high-quality dynamic 3D human images in immersive multimedia applications, efficient data compression is crucial. Unlike existing methods that focus on reducing signal-level reconstruction errors, we propose the first dynamic 3D human compression framework based on human priors. The layered coding architecture significantly enhances the perceptual quality while also supporting a variety of downstream tasks, including visual analysis and content editing. Specifically, a high-fidelity pose-driven Avatar is generated from the original frames as the basic structure layer to implicitly represent the human shape. Then, human movements between frames are parameterized via a commonly-used human prior model, i.e., the Skinned Multi-Person Linear Model (SMPL), to form the motion layer and drive the Avatar. Furthermore, the normals are also introduced as an enhancement layer to preserve fine-grained geometric details. Finally, the Avatar, SMPL parameters, and normal maps are efficiently compressed into layered semantic bitstreams. Extensive qualitative and quantitative experiments show that the proposed framework remarkably outperforms other state-of-the-art 3D codecs in terms of subjective quality with only a few bits. More notably, as the size or frame number of the 3D human sequence increases, the superiority of our framework in perceptual quality becomes more significant while saving more bitrates.

CTNet: an efficient coupled transformer network for robust hyperspectral unmixing

ABSTRACT This study introduces the coupled transformer Network (CTNet), an architecture designed to enhance the robustness and effectiveness of hyperspectral unmixing (HSU) tasks, addressing key limitations of traditional autoencoder (AE) frameworks. Traditional AEs, consisting of an encoder and a decoder, effectively learn and reconstruct low-dimensional abundance relationships from high-dimensional hyperspectral data but often struggle with spectral variability (SV) and spatial correlations, which can lead to uncertainty in the resulting abundance estimates. CTNet improves upon these limitations by incorporating a two-stream half-Siamese network with an additional encoder trained on pseudo-pure pixels, and further integrates a cross-attention module to leverage global information. This configuration not only guides the AE towards more accurate abundance estimates by directly addressing SV, but also enhances the network’s ability to capture complex spectral information. To minimize the typical reconstruction errors associated with AEs, a transcription loss constraint is applied, which preserves essential details and material-related information often lost during pixel-level reconstruction. Experimental validation on synthetic and three widely-used datasets confirms that CTNet outperforms several state-of-the-art methods, providing a more robust and effective solution for HSU challenges.

Optimizing spatial accuracy in electroencephalography reconstruction through diffuse optical tomography priors in the auditory cortex.

Diffuse optical tomography (DOT) enhances the localization accuracy of neural activity measured with electroencephalography (EEG) while preserving EEG's high temporal resolution. However, the spatial resolution of reconstructed activity diminishes for deeper neural sources. In this study, we analyzed DOT-enhanced EEG localization of neural sources modeled at depths ranging from 11-25 mm in simulations. Our findings reveal systematic biases in reconstructed depth related to DOT channel length. To address this, we developed a data-informed method for selecting DOT channels to improve the spatial accuracy of DOT-enhanced EEG reconstruction. Using our method, the average absolute reconstruction depth errors of DOT reconstruction across all depths are 0.9 ± 0.6 mm, 1.2 ± 0.9 mm, and 1.2 ± 1.1 mm under noiseless, low-level noise, and high-level noise conditions, respectively. In comparison, using fixed channel lengths resulted in errors of 2.6 ± 1.5 mm, 5.0 ± 2.6 mm, and 7.3 ± 4.5 mm under the same conditions. Consequently, our method improved the depth accuracy of DOT reconstructions and facilitated the use of more accurate spatial priors for EEG reconstructions, enhancing the overall precision of the technique.

Deep multi-sphere support vector data description based on disentangled representation learning

Deep support vector data description (Deep SVDD) combines deep mapping network and support vector data description (SVDD) to jointly optimize network connection weights and hypersphere volume. However, when the parameters of deep mapping network are set improperly, Deep SVDD may face the problem of hypersphere collapse, where all input data are mapped as the hypersphere center. To overcome the hypersphere collapse problem of Deep SVDD and improve the feature learning ability of deep mapping network, deep multi-sphere SVDD based on disentangled representation learning (DMSVDD-DRL) is proposed. DMSVDD-DRL consists of a variational autoencoder (VAE) and multiple hyperspheres. The feature representations obtained by VAE are disentangled into discriminative representations and generative representations that obey mixture t-distribution and Gaussian distribution, respectively. In the pre-training phase of DMSVDD-DRL, the network parameters and the hypersphere centers are initialized. In the training phase, the augmented data are added into the training set. The discriminative representations of both the input and augmented data are generated through the mapping network. Furthermore, multiple hyperspheres are constructed by the obtained discriminative representations in the feature space. Finally, the VAE loss of the input data, the reconstruction error of the augmented data, the augmentation loss between the input and augmented data, the average radius of the multiple hyperspheres, and the average distance from discriminative representations to their corresponding hypersphere centers are jointly minimized to obtain the optimal network connection weights and the multiple minimum volume hyperspheres. The effectiveness of the proposed DMSVDD-DRL is validated through the comparative and ablation experiments on the benchmark data sets. In addition, it is verified that DMSVDD-DRL is more robust against outliers in comparison with its related methods.

Three-dimensional surface temperature measurement using lifetime-based phosphor thermometry

Abstract In this study, we propose a three-dimensional (3D) surface temperature measurement method based on the principle of stereoscopic 3D reconstruction and the dependence of phosphorescence lifetime on temperature. A 385-nm UV(Ultraviolet) light was used as the excitation light, and two high-speed cameras were used as the detectors. The phosphor MFG (Mg4FGeO6: Mn4+) was mixed with the binder HPC and sprayed onto the tested 3D surface. The natural texture generated by the surface roughness of the phosphor coating was used as a feature for cross-correlation calculations. The digital image correlation (DIC) algorithm was used to match these feature positions in the phosphorescent images from the two cameras. The effects of the excitation angle and detecting angle were analyzed. The results indicate that the temperature measurement based on phosphorescent lifetime was not affected by the excitation and detecting angle. The method was validated on a turbine blade as an example of a 3D surface to demonstrate the capability. A comparison of the measurement results with the thermocouples proves that the current method can successfully measure the temperature on 3D surfaces with a maximum difference of 1.63°C. The spatial accuracy of the method was obtained by comparing with the measurement results of a 3D scanner, which shows that the maximum absolute error of the 3D reconstruction was 0.350 mm. The current study proposes a promising 3D surface temperature measurement method, which is expected to be widely used in gas turbine blades, Internal Combustion (IC) engine cylinders, complex curved heat exchangers, and other fields due to its non-contact measurement, low susceptibility to infrared radiation interference, high measurement accuracy, and ability to withstand harsh environments.

Image-Based 3D Reconstruction in Laparoscopy: A Review Focusing on the Quantitative Evaluation by Applying the Reconstruction Error.

Image-based 3D reconstruction enables laparoscopic applications as image-guided navigation and (autonomous) robot-assisted interventions, which require a high accuracy. The review's purpose is to present the accuracy of different techniques to label the most promising. A systematic literature search with PubMed and google scholar from 2015 to 2023 was applied by following the framework of "Review articles: purpose, process, and structure". Articles were considered when presenting a quantitative evaluation (root mean squared error and mean absolute error) of the reconstruction error (Euclidean distance between real and reconstructed surface). The search provides 995 articles, which were reduced to 48 articles after applying exclusion criteria. From these, a reconstruction error data set could be generated for the techniques of stereo vision, Shape-from-Motion, Simultaneous Localization and Mapping, deep-learning, and structured light. The reconstruction error varies from below one millimeter to higher than ten millimeters-with deep-learning and Simultaneous Localization and Mapping delivering the best results under intraoperative conditions. The high variance emerges from different experimental conditions. In conclusion, submillimeter accuracy is challenging, but promising image-based 3D reconstruction techniques could be identified. For future research, we recommend computing the reconstruction error for comparison purposes and use ex/in vivo organs as reference objects for realistic experiments.