Search

Along with the outstanding performance of the deep neural networks (DNNs), considerable research efforts have been devoted to finding ways to understand the decision of DNNs structures. In the computer vision domain, visualizing the attribution map is one of the most intuitive and understandable ways to achieve human-level interpretation. Among them, perturbation-based visualization can explain the "black box" property of the given network by optimizing perturbation masks that alter the network prediction of the target class the most. However, existing perturbation methods could make unexpected changes to network predictions after applying a perturbation mask to the input image, resulting in a loss of robustness and fidelity of the perturbation mechanisms. In this paper, we define class distortion as the unexpected changes of the network prediction during the perturbation process. To handle that, we propose a novel visual interpretation framework, Robust Perturbation, which shows robustness against the unexpected class distortion during the mask optimization. With a new cross-checking mask optimization strategy, our proposed framework perturbs the target prediction of the network while upholding the non-target predictions, providing more reliable and accurate visual explanations. We evaluate our framework on three different public datasets through extensive experiments. Furthermore, we propose a new metric for class distortion evaluation. In both quantitative and qualitative experiments, tackling the class distortion problem turns out to enhance the quality and fidelity of the visual explanation in comparison with the existing perturbation-based methods.

Since fish-eye cameras are popular in the intelligent transportation systems, accurate pedestrian detection in fish-eye images becomes more and more critical in low-speed scenarios. Usually, big data based training is the key for detectors to handle the distortion problem in fish-eye images. Especially, hard examples are more important for detectors. They have more complex features and are hard to recognize. In conventional methods, fish-eye images are collected and labeled manually. These methods are expensive and labor-intense. More importantly, these methods are still hard to collect abundant hard examples since they are rare in reality. This work proposes the Distortion Generation Network, which generates generous fish-eye images automatically using only small samples. Moreover, the Adversarial Distortion Generation Network is proposed to mine hard examples via adversarial training. These hard examples benefit detectors to be more robust to seriously distorted objects in fish-eye images. Experiments with the ETH, the KITTI, the GM-ATCI and real fish-eye datasets demonstrate that the proposed methods achieve higher accuracy than conventional methods in pedestrian detection in fish-eye images.

The signal corresponding to English speech contains a lot of redundant information and environmental interference information, which will produce a lot of distortion in the process of English speech translation signal recognition. Based on this, a large number of studies focus on encoding and processing English speech, so as to achieve high-precision speech recognition. The traditional wavelet denoising algorithm plays an obvious role in the recognition of English speech translation signals, which mainly depends on the excellent local time-frequency domain characteristics of the wavelet signal algorithm, but the traditional wavelet signal algorithm is still difficult to select the recognition threshold, and the recognition accuracy is easy to be affected. Based on this, this paper will improve the traditional wavelet denoising algorithm, abandon the single-threshold judgment of the original traditional algorithm, innovatively adopt the combination of soft threshold and hard threshold, further solve the distortion problem of the denoising algorithm in the process of English speech translation signal recognition, improve the signal-to-noise ratio of English speech recognition, and further reduce the root mean square error of the signal. Good noise reduction effect is realized, and the accuracy of speech recognition is improved. In the experiment, the algorithm is compared with the traditional algorithm based on MATLAB simulation software. The simulation results are consistent with the actual theoretical results. At the same time, the algorithm proposed in this paper has obvious advantages in the recognition accuracy of English speech translation signals, which reflects the superiority and practical value of the algorithm.

In order to solve the image blurring and distortion problem caused by underwater non-uniform and low illumination, this paper proposes an underwater image enhancement algorithm based on the Retinex theory and the Alternating Direction Method of Multipliers (ADMM). Firstly, the L component of the original image in the Lab space is extracted as the initial illumination map, and an Augmented Lagrange Multiplier (ALM) framework is constructed based on the ADMM to optimize the initial illumination map in order to obtain an accurate illumination image. In addition, the illumination map is further corrected in the luminance region with the Gamma Correction. Secondly, combined with the color constancy characteristics in the Retinex theory, the reflected image of the object is obtained. Finally, the bilateral filter is picked to suppress the underwater noise and obtain a more detailed enhanced image. The experimental results show that the underwater image enhancement algorithm can effectively solve the non-uniform illumination problem caused by natural light or artificial light source and improve the underwater image quality, thus having a better performance than other algorithms.

In this article, the ideas of post-quantum calculus and meromorphic multivalent functions are combined and some applications of these functions are discussed. We introduce a new subclass of meromorphic multivalent functions in association with Janowski domain. We investigate and study some useful geometric properties of this class of functions such as sufficiency criteria, distortion problem, growth theorem, radii of starlikeness and convexity, convex combination, and coefficient estimates for this class.

Inverse synthetic aperture radar (ISAR) imaging of near-field targets is potentially useful in some specific applications, which makes it very important to efficiently produce high-quality image of the near-field target. In this paper, the simplified target model with uniform linear motion is applied to the near-field target imaging, which overcomes the complexity of the traditional near-field imaging algorithm. According to this signal model, the method based on coordinate conversion and image interpolation combined with the range-Doppler (R-D) algorithm is proposed to correct the near-field distortion problem. Compared with the back-projection (BP) algorithm, the proposed method produces better focused ISAR images of the near-field target, and decreases the computation complexity significantly. Experimental results of the simulated data have demonstrated the effectiveness and robustness of the proposed method.

The fuzzy min–max (FMM) neural network stands as a useful model for solving pattern classification problems. FMM has many important features, such as online learning and one-pass learning. It, however, has certain limitations, especially in its learning algorithm, which consists of the expansion, overlap test, and contraction procedures. This article proposes a refined fuzzy min–max (RFMM) neural network with new procedures for tackling the key limitations of FMM. RFMM has a number of contributions. First, a new expansion procedure for overcoming the problems of overlap leniency and irregularity of hyperbox expansion is introduced . It avoids the overlap cases between hyperboxes from different classes, reducing the number of overlap cases to one (containment case). Second, a new formula that simplifies the original rules in the overlap test is proposed. It has two important features: (i) identifying the overlap leniency problem during the expansion procedure; (ii) activating the contraction procedure to eliminate the containment case. Third, a new contraction procedure for overcoming the data distortion problem and providing more accurate decision boundaries for the contracted hyperboxes is proposed. Fourth, a new prediction strategy that combines both membership function and distance measure to prevent any possible random decision-making during the test stage is proposed. The performance of RFMM is evaluated with the UCI benchmark datasets. The results demonstrate the effectiveness of the proposed modifications in making RFMM a useful model for solving pattern classification problems, as compared with other existing FMM and non-FMM classifiers.

Accurate three-dimensional (3D) intersegment joint angle measurement is an important biomechanical measure for a variety of applications. Although inertial and magnetic measurement units (IMMUs) are mobile alternatives to optical motion capture systems, the joint angle determination accuracy based on the IMMU is highly vulnerable to the magnetic distortion problem associated with magnetometer signals. This paper proposes a constraint-augmented Kalman filter (KF) for magnetometer-free 3D joint angle determination using inertial measurement unit (IMU) signals. The proposed KF uses two linear KFs consisting of a previously developed roll/pitch KF followed by a newly formulated yaw KF. In the yaw KF, the ball joint constraint provides a measurement equation for the yaw correction procedure, whereas magnetometers accomplish this procedure in most standalone algorithms. In the yaw KF, the yaw angle itself is not designated as a state because it is difficult to develop the equation in conjunction with the trigonometric functions. Instead, the sine and cosine of the yaw angle are designated as states in the form of algebraic variables. As the proposed approach does not rely on the magnetometer signals even for the 3D joint angle, it is completely free from the magnetic distortion problem. Experimental results utilizing a two-link mechanism connected by a ball joint verified that augmentation of the constraint successfully performed the role of correction. Accordingly, in magnetically undesirable conditions, the proposed IMU-based approach produced higher accuracy than the IMMU-based approaches. Furthermore, this paper investigated the residual of the constraint to provide an insight into how well each approach satisfies the kinematic constraint.

We construct a new test for correlation matrix break based on the self-normalization method. The self-normalization test has practical advantage over the existing test: easy and stable implementation; not having the singularity issue and the bandwidth selection issue of the existing test; remedying size distortion problem of the existing test under (near) singularity, serial dependence, conditional heteroscedasticity or unconditional heteroscedasticity. This advantage is demonstrated experimentally by a Monte-Carlo simulation and theoretically by showing no need for estimation of complicated covariance matrix of the sample correlations. We establish the asymptotic null distribution and consistency of the self-normalization test. We apply the correlation matrix break tests to the stock log returns of the companies of 10 largest weight of the NASDAQ 100 index and to five volatility indexes for options on individual equities.

With the development of augmented reality technology in the future, its application environment is bound to be complex and diverse, and for some extreme weather scenarios, it is very important to solve the existing problems of optical transmission display system.In addition, solving the above problems is more conducive to improving the user's perception ability of scene information in the environment with low visibility. Therefore, by analyzing the problems existing in the augmented reality display system in extreme weather, this paper proposes a set of augmented reality display system for electric navigation. First, calibrate human eyes, tracking cameras and augmented reality display devices. Aiming at the distortion problem in the display process, a four-dimensional light field calibration theory was established to reduce the display error. The camera is used to obtain the scene information, and the 3D point cloud is obtained. By denoising the 3D point cloud, the instance segmentation algorithm is used to obtain the segmentation result of the scene target. Finally, the virtual image is displayed in the scene to display the navigation information in the power operation. The system can provide clear scene information for power operators in extreme weather conditions, effectively improving work efficiency and safety.

Ultra-throw projector has been widely used in the projection application of large-FOV and short projection distance. However, the projection imaging distortion problem cannot be neglected due to the uncertain factor, such as the unsmooth factor of projection surface. In this paper, we proposed the image predistortion method based on radial basis function (RBF) image warping method to correct the projection distortion due to the unsmooth or wrinkled projection surface. The RBF image predistortion method and chief ray tracing method are used to realize a high-precision image predistortion. In addition, the distortion effect caused by the wrinkled projection surface with different depths are simulated. The distortion correction effects are also verified to less than 1% by the ray-tracing simulations.

Reverse time migration (RTM) is an ideal seismic imaging method for complex structures. However, in conventional RTM based on rectangular mesh discretization, the medium interfaces are usually distorted. Besides, reflected waves generated by the two-way wave equation can cause artifacts during imaging. To overcome these problems, a high-order finite-difference (FD) scheme and stability condition for the pseudo-space-domain first-order velocity-stress acoustic wave equation were derived, and based on the staggered-grid FD scheme, the RTM of the pseudo-space-domain acoustic wave equation was implemented. Model experiments showed that the proposed RTM of the pseudo-space-domain acoustic wave equation could systematically avoid the interface distortion problem when the velocity interfaces were considered to compute the pseudo-space-domain intervals. Moreover, this method could effectively suppress the false scattering of dipping interfaces and reflections during wavefield extrapolation, thereby reducing migration artifacts on the profile and significantly improving the quality of migration imaging.

Due to the characteristics of small-size, low cost and high resolution, the millimeter-wave synthetic aperture radar (SAR) has the unique property during the high-resolution imagery of the observations. In this article, a millimeter-wave video synthetic aperture radar (ViSAR) imaging algorithm is proposed based on the micro platforms, which can constantly produce SAR images at a high frame rate and dynamically monitor a region of interest. In the scheme, an overlapped sub-aperture strategy is applied to segment data, and two-dimensional (2D) de-chirp processing is applied to each segmented data for high resolution imaging. Finally, the Chirp-Z transform (CZT) is utilized to resolve the distortion problem of SAR image. Experiments based on simulated and measured data verify the effectiveness of the proposed algorithm.

We consider block bootstrappings for panel mean change test of the squared CUSUM test of Horvath and Huskova (J Time Ser Anal 33:631–648, 2012): the circular block bootstrapping and stationary bootstrapping. First order asymptotic null validity of the test is proved under serial and/or cross-sectional correlation. Consistency of the test under an alternative hypothesis is also proved. A Monte-Carlo experiment reveals that the existing tests of Horvath and Huskova (2012) and others have severe size distortions for serially and/or cross-sectionally correlated panels, and the block bootstrappings remedy this size distortion problem. A real data analysis illustrates the proposed method.

In this paper, the FHA model including the stray capacitance is analysed. The causes of the voltage gain distortion are studied. Limiting the working frequency is an effective method to solve the voltage gain distortion problem. A new type of the LLSC converter is proposed in the paper. It can regulate the voltage at a constant frequency. At the same time, the soft switch can be realized if the parameter design is reasonable. The proposed converter can overcome the voltage gain distortion problem and obtain the high efficiency at the light load condition. Some mathematical analyses are used to prove that the proposed converter can avoid the voltage gain distortion problems. The simulation and experimental results based on a 200W LLSC converter prototype confirm the correctness of the topology.

Wavelet Transforms (WT) have been receiving an increasing attention for the detection of damage in structures based on vibration response. The existence of the problem of boundary distortion in WT may lead to unreliable results at and around the boundaries of the structure. Previous studies have applied methods such as signal extension, zero padding and windowing to avoid border distortion. However, these methods try to avoid border distortion by distorting/reducing the original signal to prevent the distortion at the boundaries of the signal from occurring, thus altering the original signal and leading to unreliable damage identification. In this study, a WT-based method is proposed to solve boundary distortion problem. To achieve this, the mode shape difference of a structure is applied for WT decomposition. By using mode shape difference, the problem of border distortion is solved by reducing the effect of sudden change in stiffness that occurs at boundaries. A two-dimensional Continuous Wavelet Transform (CWT) is applied to decompose the difference between the damaged and the undamaged first mode shape signal. The damaged mode shape is subtracted from the undamaged mode shape to obtain the difference, and the difference is decomposed to detect and locate the damage. This method is demonstrated via numerical and experimental examples of a square steel plate. The reliability of the method is demonstrated through different damage intensities and locations. The numerical study involves applying a square plate model with two boundaries formats: all four sides fixed, and two opposite sides fixed and free. The mode shape and mode shape difference are decomposed and the coefficients of the two are compared to show the advantage of this method. The experimental example was applied to a square steel plate with all four sides fixed to verify the efficiency of the method. The numerical results showed that the problem of boundary distortion is resolved by using the mode shape difference and damage at all locations are detected, while applying mode shape provided unreliable results for both numerical models even at 25% severity. The sensitivity of the proposed method to noise is investigated by introducing various levels of signal to noise ratios and showed the detection of damage with 5% severity when mode shape difference was applied in presence of noise.

Submerged arc welding (SAW) process is extensively used in major industrial application such as wind turbine towers and ship building industries for joining of larger thickness (~10 to 50mm) of similar and dissimilar materials. However, other joining processes are not popular to weld larger thickness compared to SAW. In welding, one of the major problems is distortion, which destroys the dimensional stability of the component, a major concern in manufacturing. Simulation of larger thickness component will increase the simulation time. Therefore, an attempt has been made to evaluate a 3 - dimensional finite element analysis (FEA) using SYSWELD to study the effect of the joining of two similar shells of High Strength Low Alloy (HSLA) steel through scale down model in the ratio of 10:1. To reduce the simulation time, the two shells of steel having 400mm diameter and 2mm thickness are modelled and simulated with single pass GMAW process using eight different welding sequences as per welding procedure specification (WPS) using SYSWELD. Results show the simulated welding sequence no. 4 has a considerable amount of reduction in distortion of 1.526 mm compared to that of other sequences. Welding sequence no. 4 is then followed by actual windmill tower fabrication, which resulted in reduced distortion level of 5 mm.

In order to improve the barrel distortion of the acquired image in the surface defect detection of particleboard. In this paper, a method based on Zhang Zhengyou calibration is used to solve the camera distortion problem, so as to improve the accuracy of surface defect image processing of particleboard. Firstly, the camera was calibrated, and then the improvement of the camera correction information accuracy was judged by the correction of external parameters and reprojection error of image visualization. The internal parameter matrix and distortion coefficient of the camera were calculated accurately, and the barrel distortion of the image was corrected finally. The position of the inner corner points detected by the camera is accurate, and the reprojected points were included in the inner corner points, which improves the correction accuracy of the image to be measured. It can be clearly seen from the external parameters of visualization that the placement of the 16 sample patterns is within the range of vision. The oblique Angle deviation between the images is within 150mm. The average value of the re projection error was 0.1570 pixels calculated by the point of the camera re projection, which meets the need of correction. In conclusion, the image quality can be improved by accurately correcting the distortion of particleboard image. It lays a foundation for the surface defect extraction of particleboard.

Emotional Voice Conversion (EVC) aims to convert the emotional style of a source speech signal to a target style while preserving its content and speaker identity information. Previous emotional conversion studies do not disentangle emotional information from emotion-independent information that should be preserved, thus transforming it all in a monolithic manner and generating audio of low quality, with linguistic distortions. To address this distortion problem, we propose a novel StarGAN framework along with a two-stage training process that separates emotional features from those independent of emotion by using an autoencoder with two encoders as the generator of the Generative Adversarial Network (GAN). The proposed model achieves favourable results in both the objective evaluation and the subjective evaluation in terms of distortion, which reveals that the proposed model can effectively reduce distortion. Furthermore, in data augmentation experiments for end-to-end speech emotion recognition, the proposed StarGAN model achieves an increase of 2% in Micro-F1 and 5% in Macro-F1 compared to the baseline StarGAN model, which indicates that the proposed model is more valuable for data augmentation., Accepted by Interspeech 2021

Recent research has shown Convolutional Neural Networks (CNNs) trained in a fully-supervised fashion achieve promising performance on extreme low-light image denoising task. However, a large amount of "noisy-clean" image pairs are required to train a network, which are difficult to obtain. In this paper, we propose a compact yet effective noise model to generate synthetic noisy images for training. Especially, we address the severe color distortion problem in low-light images by identifying a novel noise component, black calibration error, as its physical origin. We prove that a small error at the sensing stage will strongly affect the following in-camera signal processing (ISP) pipeline and eventually lead to color bias. Experiment results demonstrate that the proposed model is superior in preserving perceptual quality and achieves state-of-the-art performance among existing noise synthesis methods.

An image can be decomposed into two parts: the basic content and details, which usually correspond to the low-frequency and high-frequency information of the image. For a hazy image, these two parts are often affected by haze in different levels, e.g. , high-frequency parts are often affected more serious than low-frequency parts. In this paper, we approach the single image dehazing problem as two restoration problems of recovering basic content and image details, and propose a Dual-Path Recurrent Network (DPRN) to simultaneously tackle these two problems. Specifically, the core structure of DPRN is a dual-path block, which uses two parallel branches to learn the characteristics of the basic content and details of hazy images. Each branch consists of several Convolutional LSTM blocks and convolution layers. Moreover, a parallel interaction function is incorporated into the dual-path block, thus enables each branch to dynamically fuse the intermediate features of both the basic content and image details. In this way, both branches can benefit from each other, and recover the basic content and image details alternately, therefore alleviating the color distortion problem in the dehazing process. Experimental results show that the proposed DPRN outperforms state-of-the-art image dehazing methods in terms of both quantitative accuracy and qualitative visual effect.

To mitigate the rabbit hole effect in recommendations, conventional distribution-aware recommendation systems aim to ensure that a user’s prior interest areas are reflected in the recommendations that the system makes. For example, a user who historically prefers comedies to dramas by 2:1 should see a similar ratio in recommended movies. Such approaches have proven to be an important building block for recommendation tasks. However, existing distribution-aware approaches enforce that the target taste distribution should exactly match a user’s prior interests (typically revealed through training data), based on the assumption that users’ taste distribution is fundamentally static. This assumption can lead to large estimation errors. We empirically identify this taste distortion problem through a data-driven study over multiple datasets. We show how taste preferences dynamically shift and how the design of a calibration mechanism should be designed with these shifts in mind. We further demonstrate how to incorporate these shifts into a taste enhanced calibrated recommender system, which results in simultaneously mitigated both the rabbit hole effect and taste distortion problem.

Slope failure induced by sublevel caving mining is a progressive process, resulting in the large deformation and displacement of rock masses in the slope. Numerical methods are widely used to investigate the above phenomenon. However, conventional numerical methods have difficulties when simulating the process of progressive slope failure. For example, the discrete element method (DEM) for block systems is computationally expensive and possibly fails for large-scale and complex slope models, while the finite difference method (FDM) has a mesh distortion problem when simulating progressive slope failure. To address the above problems, this paper presents a finite difference modeling method using the adaptive local remeshing technique (LREM) to investigate the progressive slope failure induced by sublevel caving mining. In the proposed LREM, (1) the zone of the distorted mesh is adaptively identified, and the landslide body is removed, (2) the updated mesh is regenerated by the local remeshing, and the physical field variables of the original computational model are transferred to the regenerated computational model. The novelty of the proposed method is that (1) compared with the DEM for block systems, the proposed LREM is capable of modeling the progressive slope failure in large-scale rock slopes, (2) the proposed method is able to address the problem of mesh distortion in conventional FDM modeling, and (3) compared with the errors induced by the frequent updating of the mesh of the entire model, the adaptive local remeshing technique effectively reduces calculation errors. To evaluate the effectiveness of the proposed LREM, it is first used to investigate the failure of a simplified slope induced by sublevel caving mining. Moreover, the proposed LREM is applied in a real case, i.e., to investigate the progressive slope failure induced by sublevel caving mining in Yanqianshan Iron Mine.

Underwater image restoration has attracted the attention of many experts due to the demand of underwater rescue, archaeology, resource exploration et al. Water scattering and absorption hinder the linear propagation of light in water, resulting the image degradation. The degraded image visibility can be improved by the image restoration method. Many image restoration methods have been proposed by the experts for underwater images restoration. The dark channel prior method is a classical method for image dehazing, and it also can be used for underwater image restoration. However, the underwater image restored by this method still has the color distortion problem due to the water absorption, and the less visibility due to the inaccuracy transmission estimation. To overcome this problems, we propose an image restoration method based on improved dark channel prior to repair the image degraded by underwater scattering media. The quadtree decomposition is used to find the high intensity region for waterlight estimation. The transmission is estimated by the improved dark channel prior. The underwater degraded image can be restored by our proposed method without color distortion. This method will be helpful to the underwater rescue and resource exploration et al.

Friction welding is a solid-state welding process that generates heat through mechanical friction between workpieces in relative motion to one another, with the addition of a lateral force called "upset" to plastically displace and fuse the materials. One of the key problems during modeling solid-state welding processes is the implementation of the friction model. Coulomb friction model was proposed by researchers for computational solid mechanics (CSMs) modeling of the process, however this law is based on the sliding motion. Norton friction model is also proposed by others, while simplifications and limitations are caused in coincidence with reality. In this paper, Norton friction model is modified to be employed in a computational solid mechanics model of inertia welding. A continuous remeshing technique is used to avoid the mesh distortion problem. Consequently, successful prediction of the temperature distribution, thermal history, equivalent plastic deformation, axial shortening and stress distribution is made. The comparisons between the results of this study and the literature showed that implementing the proposed methodology leads to achieving high accuracy results.

The recycling of marine exploration equipment after it has surfaced is greatly affected by sea state. In order to estimate the sea state in real time, this paper proposes a method for measuring wave elevation, which modifies the integrated results of GNSS/SINS in the up direction by virtual horizontal lines to extract wave fluctuation information. From these wave information, the significant wave heights (SWH) can be calculated as the only input parameter of P-M spectrum, and a series of wave height data can be further simulated. When the GNSS is interrupted due to severe sea state, the simulated data can be integrated with the SINS to deal with the data distortion problem. The simulation results show that the application of wave spectrum in the GNSS intermittent situation has obvious improvement effect and important significance.

In this work we study the metric distortion problem in voting theory under a limited amount of ordinal information. Our primary contribution is threefold. First, we consider mechanisms which perform a sequence of pairwise comparisons between candidates. We show that a widely-popular deterministic mechanism employed in most knockout phases yields distortion \(\mathcal {O}(\log m)\) while eliciting only \(m-1\) out of \(\varTheta (m^2)\) possible pairwise comparisons, where m represents the number of candidates. We also provide a matching lower bound on its distortion. In contrast, any mechanism which performs fewer than \(m-1\) pairwise comparisons has unbounded distortion. Moreover, we study the power of deterministic mechanisms under incomplete rankings. Most notably, when every agent provides her k-top preferences we show an upper bound of \(6 m/k + 1\) on the distortion, for any \(k \in \{1, 2, \dots , m\}\), substantially improving over the previous bound of 12m/k recently established by Kempe [25, 26]. Finally, we are concerned with the sample complexity required to ensure near-optimal distortion with high probability. Our main contribution is to show that a random sample of \(\varTheta (m/\epsilon ^2)\) voters suffices to guarantee distortion \(3 + \epsilon \) with high probability, for any sufficiently small \(\epsilon > 0\). This result is based on analyzing the sensitivity of the deterministic mechanism introduced by Gkatzelis, Halpern, and Shah [22].

It is very challenging to reconstruct a high dynamic range (HDR) from a low dynamic range (LDR) image as an ill-posed problem. This paper proposes a luminance attentive network named LANet for HDR reconstruction from a single LDR image. Our method is based on two fundamental observations: (1) HDR images stored in relative luminance are scale-invariant, which means the HDR images will hold the same information when multiplied by any positive real number. Based on this observation, we propose a novel normalization method called " HDR calibration " for HDR images stored in relative luminance, calibrating HDR images into a similar luminance scale according to the LDR images. (2) The main difference between HDR images and LDR images is in under-/over-exposed areas, especially those highlighted. Following this observation, we propose a luminance attention module with a two-stream structure for LANet to pay more attention to the under-/over-exposed areas. In addition, we propose an extended network called panoLANet for HDR panorama reconstruction from an LDR panorama and build a dualnet structure for panoLANet to solve the distortion problem caused by the equirectangular panorama. Extensive experiments show that our proposed approach LANet can reconstruct visually convincing HDR images and demonstrate its superiority over state-of-the-art approaches in terms of all metrics in inverse tone mapping. The image-based lighting application with our proposed panoLANet also demonstrates that our method can simulate natural scene lighting using only LDR panorama. Our source code is available at https://github.com/LWT3437/LANet.

With great growth in the volume of chemical databases, machine learning receives significant attention from various scientific communities for efficient high-throughput screening of molecular properties and drug discovery on the millions of chemical compounds. In particular, graph neural networks (GNNs) have been widely studied in chemistry-related fields because a molecule is natively represented as a mathematical graph. In GNNs for the graph-level analysis, a global operation called readout is applied after node embedding to generate a graph-level embedding that represents characteristics of the whole graph. However, commonly used readouts frequently distort scale information of the graph and consequently degrade the prediction accuracy of GNNs. This problem becomes more serious in molecular machine learning because molecules have many important scale information (e.g., molecular weight and total energy). In this paper, we investigate this scale distortion problem in GNNs caused by the readouts for the first time and propose an efficient solution with a new attention-based readout. In the experiments, the proposed readout outperformed commonly used readouts on various GNN architectures.

This paper applies a body force model developed recently to investigate the interaction between total temperature distortion and a multistage fan. The off-design performance of the fan shows the reasonable predicting accuracy and supports the present model is applicable for high-speed multistage machines. The transfer behaviors of 90° steady-state circumferential total temperature distortion as well as combined total pressure and total temperature distortion in the multistage environment are captured successfully by the model. The mechanism of the phase shift of the high temperature sector is discussed by the model to advance the understanding of the total temperature distortion problem. The results reveal that the large-scale flow feature of total temperature distortion in the multistage environment can be capably quantified by the present body force model with the acceptable computational consumption.

The color and texture data used in traditional digitalization fields based on three primaries (RGB) have multiple non-linear editing, which causes color distortion and unable to meet demand of protection, utilization of cultural heritage. Whereas the digitalization approaches based on multi-spectral has aroused widespread interest by uniquely objectively expressing the material color of object surface. This project intends to use multi-spectral method (multi-spectral reflection rate and transmission rate of objects) to objectively express the material color of surface, beyond the limit of three primaries (RGB) expression, establishes equipment-free, environment-free color acquisition method and expression model combining compressive sensing technology. High-fidelity virtual reappearance algorithm based on multi-spectral data is studied, which is based on existing high-realistic rendering technology. The compressive sensing technology is utilized to speed up virtual appearance. The above approaches could resolve the distortion problem of the application fields requiring high color fidelity. A multi-spectral software and hardware prototype system for data acquisition, modeling, high-fidelity virtual appearance will be implemented based on above algorithm.

The circulating current (CC) occurs among the paralleled converters due to the inconsistency of impedance parameters, carrier phases, switching frequencies and so on, and for Vienna rectifier, which contains the current distortion problem, will bring worse input current. Therefore, in this paper, an integrated control method is proposed to suppress the CC and current distortion. Specifically, the modeling of CC in paralleled Vienna rectifiers is established, on the basis of normal zero-sequence circulating current (ZSCC) control, a feedforward control is added to eliminate the ZSCC spikes. For the current distortion, the corresponding analysis and solution in space vector modulation (SVM) is conducted, to facilitate the implementation, the zero-sequence component injected by the equivalent carrier-based PWM method is deduced. Finally, the feasibility of the proposed integrated control strategy is validated by simulation and experiment.

Predicting the potential target customers for a product is essential. However, traditional recommender systems typically aim to optimize an engagement metric without considering the overall distribution of target customers, thereby leading to serious distortion problems. In this paper, we conduct a data-driven study to reveal several distortions that arise from conventional recommenders. Toward overcoming these issues, we propose a target customer re-ranking algorithm to adjust the population distribution and composition in the Top-k target customers of an item while maintaining recommendation quality. By applying this proposed algorithm onto a real-world dataset, we find the proposed method can effectively make the class distribution of items’ target customers close to the desired distribution, thereby mitigating distortion.

In order to reduce the times of changing physical target in shooting training and competition system, count the scores of shots accurately, the bullet hole images of each shot are collected and analyzed, the new bullet holes are count by the subtraction method, dazzle light problems are solved by locally self-adaptive threshold method, hole shape distortion problem is solved with corrosion and expansion method, hole positioning problem is solved by using the inner and outer circle median method. Experimental results show that these methods can obtain higher accuracy and image quality.

Remote sensing technology has been widely used for monitoring crop fields and other agricultural applications. However, the clarity of remote sensing images is often affected by clouds and chaotic media in the atmosphere. Image dehazing can be achieved through the dark channel prior method (DCP), but there is always a brightness distortion problem after image dehazing. To address the problem, this study proposed an improved image dehazing approach based on the DCP method and determined optimal enhancement parameters. Four evaluation indices, including mean square error (MSE), peak signal to noise ratio (PSNR), average gradient and program running time, were first calculated to evaluate the quality of enhanced images. An example image was dehazed by the DCP method initially using the four indices to determine optimal dehazing parameters. Results showed that image enhancement achieved the best effect when the dark channel window size Ω(x) is 5, atmospheric light A is 215/255, and the lower limit t0 of transmission factor t(x) is 0.1. Next, these indices were applied to evaluate the enhancement methods used in this research. The logarithmic enhancement method was finally selected as the optimal method with the base number (1 + r) = 11 and enhancement parameter m = 0.5. To verify the effectiveness of the selected method, 50 airborne images from a consumer-grade camera flown by an agricultural aircraft were used to evaluate the improved method. Both the original and the enhanced images after dehazing were mosaicked by Adobe Photoshop software. The mosaicked images before and after image dehazing were compared. Results showed that the mosaicked image without dehazing had an entropy of 6.359 and an average gradient of 6.513. In comparison, the mosaicked image with dehazing had an entropy of 6.668 and an average gradient of 11.305, which were 4.86% and 73.58% higher than the respective values for the mosaicked image without dehazing. These results indicate that the proposed method in this study is effective and can be applied to dehaze remote sensing images.

In 1994 E. Odell and Th. Schlumprecht showed that the Hilbert space (` 2 ) was arbitrarily distortable.However, we do not have a defined sequence of norms that would perform this distortion. G. An-droulakis and the author defined a class of norms equivalent to the Hilbert space norm. These normsare candidate sequences of norms that might distort ` 2 . These norms are recursively defined andcalculations with these norms are quite difficult. Here we present a tool, a Python program, thatcalculates the norms defined by G. Androulakis and the author.

Teaching computer programs to play games through machine learning has been an important way to achieve better artificial intelligence (AI) in a variety of real-world applications. Monte Carlo Tree Search (MCTS) is one of the key AI techniques developed recently that enabled AlphaGo to defeat a legendary professional Go player. What makes MCTS particularly attractive is that it only understands the basic rules of the game and does not rely on expert-level knowledge. Researchers thus expect that MCTS can be applied to other complex AI problems where domain-specific expert-level knowledge is not yet available. So far there are very few analytic studies in the literature. In this paper, our goal is to develop analytic studies of MCTS to build a more fundamental understanding of the algorithms and their applicability in complex AI problems. We start with a simple version of MCTS, called random playout search (RPS), to play Tic-Tac-Toe, and find that RPS may fail to discover the correct moves even in a very simple game position of Tic-Tac-Toe. Both the probability analysis and simulation have confirmed our discovery. We continue our studies with the full version of MCTS to play Gomoku and find that while MCTS has shown great success in playing more sophisticated games like Go, it is not effective to address the problem of sudden death/win. The main reason that MCTS often fails to detect sudden death/win lies in the random playout search nature of MCTS, which leads to prediction distortion. Therefore, although MCTS in theory converges to the optimal minimax search, with real world computational resource constraints, MCTS has to rely on RPS as an important step in its search process, therefore suffering from the same fundamental prediction distortion problem as RPS does. By examining the detailed statistics of the scores in MCTS, we investigate a variety of scenarios where MCTS fails to detect sudden death/win. Finally, we propose an improved MCTS algorithm by incorporating minimax search to overcome prediction distortion. Our simulation has confirmed the effectiveness of the proposed algorithm. We provide an estimate of the additional computational costs of this new algorithm to detect sudden death/win and discuss heuristic strategies to further reduce the search complexity.

Image distortion is inevitable when an image is captured through a lens. While the digital image measurement technique is getting popular, image distortion problem can result in significant error. A new distortion correction method is proposed in this study. The proposed method is based on the fact that a flat surface should keep flat when it is measured using three-dimensional (3D) digital image measurement technique. The 3D digital image measurement technique adopted in this research is the simplified 3D digital image correlation (DIC) method. Because radial distortion has a more noticeable influence than other types of distortions, this method deals only with radial distortion. A few experiments are carried out in this study to verify the correctness of this method and its accuracy. Both simulated data and actual image data are adopted in these experiments. The results show that this method can achieve a good accuracy. The standard deviations caused by random errors are about the same order as the random errors. It also shows that this method is suitable for both large and small distortion conditions.

In most large field of view (FOV) observations, the distortion problem is inevitably and significantly more serious than in small FOV ones. In the circumstances, many traditional star ident...

Haze degrades visual information of remotely sensed images. Therefore, haze removal is a demanding and significant task for visual multispectral information improvement. The existing haze removal techniques utilize different restrictions and before restoring hazy images in an efficient manner. The review of existing haze removal methods demonstrates that the haze-free images suffer from colour distortion and halo artefacts problems. To solve these issues, an improved restoration model based dark channel prior is proposed in this paper. The proposed technique has redefined transmission map, with the aim to reduce the colour distortion problem. The modified joint trilateral filter is also utilized to improve the coarse estimated atmospheric veil. The experimental results reveal that the proposed approach provides visually significant haze-free images and also preserves the significant detail.

The structural distortion problem of large thin-walled aerospace components has roused much concern on more agile, digitized and cost-efficient precision manufacturing techniques, in particular, designing suitable machining plans according to the real shape of workpieces. To improve the machining accuracy of large thin-walled aerospace components, a methodology of three-dimensional machining allowance modeling and analysis is proposed in this paper. Firstly, the fundamental principles of 3D models matching and alignment between the casting blank and the design part are presented. And by datum transformation from virtual design references to real-world references on casting blanks which are accessible more easily, measuring, positioning and machining the casting blank will be more convenient to realize. Furthermore, the technical character of a large thin-walled aerospace component is considered and a technical framework for 3D machining allowance analysis is given. Finally, the proposed methodology is validated by an experimental study using a typical large thin-walled aerospace component of a hypersonic vehicle. Improvements in accuracy and efficiency performance are realized in comparison with traditional methods.

Fundamentally, heat is needed during friction stir welding (FSW) for joining materials together. This heat is mainly produced by the friction force which is a part of the force that is applied to the tool. Many researchers have tried to use finite element modelling for investigating the thermal and the material flow behaviour; however, the process large plastic deformation causes mesh distortion. To illustrate, the excessive mesh distortion should be considered as one of the key problems happening during the modelling of the process. In this paper, the thermomechanical behaviour of FSW is investigated using finite element modelling. The smoothed-particle hydrodynamics (SPH) method is used to solve the mesh distortion problem. The outcomes indicated that the temperature of the welding is around 524 °C that is less than the welding material melting point. Finally, the outcomes are matched with some scientific papers which this issue confirms the accuracy of the simulated model.

The dual active bridge (DAB) topology is frequently preferred in industry as a dc–dc converter with various current modulation methods, such as phase shift modulation (PSM), trapezoidal modulation (TZM), triangular modulation (TRM), and hybrid methods. The TZM and TZM–TRM hybrid methods can solve the current circulation problem of PSM, extend the zero voltage switching region to the full operating range, and thereby improve the light load efficiency. However, all these methods have been developed for a dc–dc converter, and cannot be directly applied to a single-stage DAB-based ac–dc converter due to the low input voltage close to the zero crossing points of the grid, otherwise significant distortion is created. This paper analyzes the line current distortion related with the modulation for a single-stage DAB-based ac–dc converter, and a novel hybrid current modulation (HCM) method that solves the distortion problem is proposed. However, since the HCM method increases the rms current as side effect, the improved version of the iHCM method is developed and the distortion-free current without increasing the rms current is obtained. The results obtained from the theoretical analysis are verified by an experimental prototype of 250 W.

Computational fluid dynamics (CFD)-based numerical wave tanks are valuable tools for the development and evaluation of energy maximising control systems for wave energy converters (WECs). However, the exaggerated body motion amplitude, which can be induced by the energy maximising control system, challenges the commonly applied mesh morphing method in CFD, due to the resulting mesh distortion and subsequent numerical instability. A more advanced mesh motion method is the overset grid method, which can inherently handle large-amplitude body motions and has recently become freely available in the open-source CFD software OpenFOAM. The overset grid method can, therefore, potentially eliminate the mesh distortion problem, hindering the simulation of WECs under controlled conditions. To evaluate the capability of the overset grid method for control studies of WECs in an OpenFOAM numerical wave tank, this paper presents a detailed comparison of the overset grid and mesh morphing methods, considering five test cases of increasing complexity. The test cases range from a static equilibrium test to the modelling of a controlled WEC, and good agreement is demonstrated between the two mesh motion methods, except for the case of the controlled WEC, when the device motion becomes large, and the mesh morphing simulation crashes. The runtimes for overset grid simulations are observed to be approximately double the time required for the mesh morphing simulations.

Capacity of grid participants such as photovoltaic panel, wind power and electric vehicles has shown rapid growth, while the technology in the industrial sector has continued to innovate, and many types of power load characteristics have undergone tremendous changes. This paper proposes an improved worker intelligence load forecasting method based on trajectory clustering. The method clusters different fluctuation patterns of daily load curves, summarizes the influencing factors, and improves the intelligent prediction method. It is suitable for load forecasting in areas where load characteristics are not significantly consistent and where external stimuli are significantly affected. Solve the distortion problem of traditional load forecasting for continuous prediction of new energy and random power load. The calculations are carried out for areas with industrial agglomeration and new energy capacity. The results show that the method can accurately reflect the future load trend from the perspective of feature correction.

This paper investigates the speed compensation of range profile for space target inverse synthetic aperture radar (ISAR) imaging. By establishing an ISAR echo model, the impact of the fast-moving space target on the performance of range distance pulse compression has been analyzed and we find out that the fast-moving target will distort the range profile. To solve the distortion problem, a range profile compensation algorithm based on Chirplet transform has been proposed. The new algorithm can eliminate the influence of the speed on the range profile and improve the imaging quality. The orbit of the target are simulated by using STK software. Simulated imaging results are given at the end of the paper to verify the effectiveness and efficiency of the proposed algorithm.

Muscle synergy is the association of several muscle groups contracting simultaneously as a single functional unit. Recently, the synergy has been emphasized in neurophysiological studies to utilize it as an evaluation index for stroke patients. In this paper, we proposed novel method for quantitatively comparing muscle synergy of normal people and the stroke patient for evaluating severity of the stroke. The proposed method uses NNR (Non-Negative Matrix Reconstruction) algorithm for reconstructing muscle weight matrix W. The proposed method synchronizes the time sequence of EMG data of the stroke patient with the normal people based on working phases. The method improves the distortion problem of NNR algorithm through the synchronization of time sequence.

This paper studies the strongly-coupled inductive power transfer (IPT) system, which reveals that the high order harmonic distortions in the input and output currents become significant when the coupling coefficient is relatively high. The common used series-series (SS) compensation circuit is analyzed in this paper. The theoretical model of the circuit is presented for different frequencies. The analysis results show that the transfer power in the system closely relates to the harmonic components. Specifically, the distortion problem becomes more severe with an increasing coupling coefficient. In addition, this digest introduces an effective method to design the circuit parameters in order to reduce the distortion. At last, a 1.8 kW prototype is designed and implemented, and the measurement shows a coupling coefficient as high as 0.75. Experimental results validate the relatively large high order harmonic distortion problem as proposed. Meanwhile, with proper parameter design, the maximum current distortion is limited within 37%, which is acceptable for the real application.

In this paper, the servo system mechanical drive is analyzed, flexible beam vibration will cause end point trajectory distortion. Simplify the flexible beam load as a second-order system, which is a low frequency light damped vibration system, the ZVD input shaper which is very suitable for vibration suppression of low frequency light damping system is adopted. Due to the different vibration frequencies and damping coefficients of the X-Y platform flexible load, the X-axis and Y-axis input shapers cause different command lags, which can cause trajectory distortion. The adaptive feedforward controller based on phase angle error is used to solve trajectory distortion problem. Combined with the input shaper and adaptive feedforward control based on phase error, it can solve the vibration suppression of X-axis and Y-axis flexible load with different vibration characteristics, and can guarantee the track accuracy. Finally, the effectiveness of the algorithms is verified by simulation.