Conventional Interpolation Research Articles

Accurate and Scalable Gaussian Processes for Fine-Grained Air Quality Inference

Air pollution is a global problem and severely impacts human health. Fine-grained air quality (AQ) monitoring is important in mitigating air pollution. However, existing AQ station deployments are sparse. Conventional interpolation techniques fail to learn the complex AQ phenomena. Physics-based models require domain knowledge and pollution source data for AQ modeling. In this work, we propose a Gaussian processes based approach for estimating AQ. The important features of our approach are: a) a non-stationary (NS) kernel to allow input depended smoothness of fit; b) a Hamming distance-based kernel for categorical features; and c) a locally periodic kernel to capture temporal periodicity. We leverage batch-wise training to scale our approach to a large amount of data. Our approach outperforms the conventional baselines and a state-of-the-art neural attention-based approach.

Combining Numerical Simulation and Deep Learning for Landslide Displacement Prediction: An Attempt to Expand the Deep Learning Dataset

Effective landslide hazard prevention requires accurate landslide prediction models, and the data-driven approaches based on deep learning models are gradually becoming a hot research topic. When training deep learning models, it is always preferable to have a large dataset, while most available landslide monitoring data are limited. For data missing or data sparseness problems, conventional interpolation methods based on mathematical knowledge lack mechanism interpretability. This paper proposes that numerical simulations can be used to expand the deep learning dataset we need. Taking the Jiuxianping landslide in the Three Gorges Reservoir Area (TGRA) as the geological background, a finite element numerical model was established, and the landslide displacement time series data were solved considering the boundary conditions of reservoir water level change and precipitation. Next, based on three metrics: Euclidean distance, cosine similarity, and dynamic time warping (DTW) distance, the time series similarity between the displacement data obtained from simulation and data obtained from actual monitoring were verified. Finally, the combined deep learning model was built to predict the displacement of the Jiuxianping landslide. The model was trained on both the simulated and monitored datasets and tested by the last 12 monitored data points. Prediction results with the testing set showed that the models trained using the expanded training set from numerical simulations exhibited lower prediction errors, and the errors had a more concentrated distribution. The results suggest that this landslide displacement prediction method combining numerical simulation and deep learning can solve the problem of inadequate datasets due to low monitoring frequency, as well as provide an interpretation of the physical mechanism for data vacancy filling.

Tensor interpolation in virtual manufacturing chains for fiber reinforced composites

Preeminent weight-specific mechanical properties predestine fiber-reinforced plastics for application in structural components. Virtual manufacturing chains (CAE-chains) capture a multi-step process sequence by interlinking models with corresponding constitutive laws for each process step. In common numerical solving techniques, spatial discretization of governing equations yields discrete solutions. Since mesh type and fineness are usually problem-specific, mesh-to-mesh mapping must be embedded in the data interfaces between the individual simulation steps. To receive meaningful data, the underlying averaging and interpolation schemes between non-congruent meshes must be mathematically and physically consistent. In engineering applications, interpolation of tensors is usually carried out component-based, treating each component as an independent scalar-field and thus neglecting the tensorial character. The work at hand gives an overview of sophisticated techniques that preserve specific tensor characteristics in interpolation. Compared to available approaches, an enhanced decomposition-based interpolation method is proposed, allowing for a generalization with more than two basic values. In the context of process simulation for short fiber-reinforced injection molding, the influence of the interpolation techniques is evaluated in a three-stage approach with increasing proximity to application. Firstly, mathematical examples demonstrate that component-based interpolation can result in erroneous, non-monotonous tensor characteristics. Secondly, an analytically resolvable problem is derived, and the solution is compared to reconstructed tensors from different interpolation schemes. Thirdly, a numerical CFD-simulation is conducted. Field recovery via interpolation is performed and systematical errors are statistically evaluated. The results reveal significant biases for certain tensor characteristics induced by conventional component-wise interpolation. Overall, a reduction in systematic interpolation errors is achieved by the proposed decomposition-based interpolation.

Simplified estimation modeling of land surface solar irradiation: A comparative study in Australia and China

Solar irradiation maps are fundamental geospatial datasets that have been used in a variety of research fields. However, it is difficult to estimate the continuous distribution of solar irradiation over large areas accurately by using conventional interpolation or extrapolation methods based on only a few observation stations. To tackle this problem, this study proposed a method to estimate spatially continuous land surface solar irradiation based on four machine learning models, namely, Gradient Boosting Machine (GBM), Random Forest (RF), Support Vector Regression (SVR), and Multilayer Perceptron (MLP). Clear-sky solar irradiation data were computed based on time and location, cloud optical thickness (COT) and aerosol optical thickness (AOT) that significantly influence solar irradiation were retrieved from Himawari-8 meteorological satellite images, and land surface solar irradiation data were obtained from observation stations for training and evaluation. To explore the weather effects on land surface solar irradiation, air temperatures, humidity, wind, and atmospheric pressure were also quantified and integrated into the models. As a comparative study, this study collected six-year historical data and estimated solar distribution at a 5-km spatial resolution in Australia and China. Based on the coefficient of determination (R2), normalized Root Mean Square Error (nRMSE), normalized mean bias error (nMBE), and consumption of time (t), the results show that GBM achieved the highest accuracy with R2>0.7 at all stations, followed by RF, SVR, and MLP. It suggests that the proposed method can provide an accurate and reliable estimation of land surface solar irradiation, compared with the theoretical solar irradiation without the obstacle of the atmosphere. The annual solar distribution maps created by the built methods indicate that the proposed method is simple and effective for large geographical regions and can be used worldwide when similar datasets are obtained.

ENHANCED SUPER RESOLUTION FOR REMOTE SENSING IMAGERIES

Abstract. Single image super resolution (SISR) technology has been attracted much attention from remote sensing community due to its proven potentials in remote sensing applications. Existing SISR techniques varying from conventional interpolation methods to different network architectures. Generative adversarial networks (GANs) are one of the latest network architectures proven a greater potential as a SISR method whereas least attention has been given by the remote sensing community. Several studies have already been carried out on this context. However, yet there is no generalized GAN based approach to super resolve remote sensing imageries. Therefore, this study investigated the potentials of enhanced super resolution generative adversarial (ESRGAN) model to super resolve very high to medium resolution images from high to coarse resolution images for remote sensing applications. Two models were trained and Worldview-3 (WV3) images used as for very high resolution images. Whereas, down sampled WV3 and Sentinel-2(S2) were used as low resolution counterparts. Model performances were qualitatively and quantitatively analysed using standard metrics such as PSNR, SSIM, UIQI, CC, SAM, SID. Evaluation results emphasised super resolved images were preserved the original quality of the satellite images to a greater extent while improving its ground resolution.

Through-Plane Super-Resolution With Autoencoders in Diffusion Magnetic Resonance Imaging of the Developing Human Brain.

Fetal brain diffusion magnetic resonance images (MRI) are often acquired with a lower through-plane than in-plane resolution. This anisotropy is often overcome by classical upsampling methods such as linear or cubic interpolation. In this work, we employ an unsupervised learning algorithm using an autoencoder neural network for single-image through-plane super-resolution by leveraging a large amount of data. Our framework, which can also be used for slice outliers replacement, overperformed conventional interpolations quantitatively and qualitatively on pre-term newborns of the developing Human Connectome Project. The evaluation was performed on both the original diffusion-weighted signal and the estimated diffusion tensor maps. A byproduct of our autoencoder was its ability to act as a denoiser. The network was able to generalize fetal data with different levels of motions and we qualitatively showed its consistency, hence supporting the relevance of pre-term datasets to improve the processing of fetal brain images.

A 9-bit 500-MS/s 2-bit/cycle SAR ADC With Error-Tolerant Interpolation Technique

This article presents a 9-bit 500-MS/s 2-bit/cycle successive approximation register (SAR) analog-to-digital converter (ADC) with an error-tolerant interpolation technique. The proposed interpolation technique uses flip-flops to implement a 2-bit/cycle operation in the SAR ADC. By taking advantage of the metastable region of the flip-flop, the proposed interpolator can defer the bit decision when a decision error occurs with a high probability. Because the SAR ADC approximates the signal range step by step, the deferred decisions proceed to the next conversion cycles without any increase in quantization noise. The deferring-decision characteristic increases the error tolerance in the presence of comparator mismatches and increases the inherent linearity of the interpolation technique compared to conventional latch interpolation. A prototype ADC was designed using the 28-nm CMOS technology to verify the effectiveness of the proposed interpolation technique. The measured signal-to-noise-and-distortion ratio (SNDR) and spurious-free dynamic range (SFDR) at the Nyquist rate are 50.6 and 61.4 dB, respectively. The power consumption is 1.87 mW at a sampling frequency of 500 MS/s. The proposed ADC achieves a Walden figure of merit (FoM) of 13.5 fJ/conversion-step.

Symmetric quadratic tetration interpolation using forward and backward operation combination

The existed interpolation method, based on the second-order tetration polynomial, has the asymmetric property. The interpolation results, for each considering region, give individual characteristics. Although the interpolation performance has been better than the conventional methods, the symmetric property for signal interpolation is also necessary. In this paper, we propose the symmetric interpolation formulas derived from the second-order tetration polynomial. The combination of the forward and backward operations was employed to construct two types of the symmetric interpolation. Several resolutions of the fundamental signals were used to evaluate the signal reconstruction performance. The results show that the proposed interpolations can be used to reconstruct the fundamental signal and its peak signal to noise ratio (PSNR) is superior to the conventional interpolation methods, except the cubic spline interpolation for the sine wave signal. However, the visual results show that it has a small difference. Moreover, our proposed interpolations converge to the steady-state faster than the cubic spline interpolation. In addition, the option number increasing will reinforce their sensitivity.

PARTITIONING AROUND MEDOIDS CLUSTERING AND RANDOM FOREST CLASSIFICATION FOR GIS-INFORMED IMPUTATION OF FLUORIDE CONCENTRATION DATA.

Community water fluoridation is an important component of oral health promotion, as fluoride exposure is a well-documented dental caries-preventive agent. Direct measurements of domestic water fluoride content provide valuable information regarding individuals' fluoride exposure and thus caries risk; however, they are logistically challenging to carry out at a large scale in oral health research. This article describes the development and evaluation of a novel method for the imputation of missing domestic water fluoride concentration data informed by spatial autocorrelation. The context is a state-wide epidemiologic study of pediatric oral health in North Carolina, where domestic water fluoride concentration information was missing for approximately 75% of study participants with clinical data on dental caries. A new machine-learning-based imputation method that combines partitioning around medoids clustering and random forest classification (PAMRF) is developed and implemented. Imputed values are filtered according to allowable error rates or target sample size, depending on the requirements of each application. In leave-one-out cross-validation and simulation studies, PAMRF outperforms four existing imputation approaches-two conventional spatial interpolation methods (i.e., inverse-distance weighting, IDW and universal kriging, UK) and two supervised learning methods (k-nearest neighbors, KNN and classification and regression trees, CART). The inclusion of multiply imputed values in the estimation of the association between fluoride concentration and dental caries prevalence resulted in essentially no change in PAMRF estimates but substantial gains in precision due to larger effective sample size. PAMRF is a powerful new method for the imputation of missing fluoride values where geographical information exists.

Modified recurrent equation-based cubic spline interpolation for missing data recovery in phasor measurement unit (PMU)

Background: Smart grid systems require high-quality phasor measurement unit (PMU) data for proper operation, control, and decision-making. Missing PMU data may lead to improper actions or even blackouts. While the conventional cubic interpolation methods based on the solution of a set of linear equations to solve for the cubic spline coefficients have been applied by many researchers for interpolation of missing data, the computational complexity increases non-linearly with increasing data size. Methods: In this work, a modified recurrent equation-based cubic spline interpolation procedure for recovering missing PMU data is proposed. The recurrent equation-based method makes the computations of spline constants simpler. Using PMU data from the State Load Despatch Center (SLDC) in Madhya Pradesh, India, a comparison of the root mean square error (RMSE) values and time of calculation (ToC) is calculated for both methods. Results: The modified recurrent relation method could retrieve missing values 10 times faster when compared to the conventional cubic interpolation method based on the solution of a set of linear equations. The RMSE values have shown the proposed method is effective even for special cases of missing values (edges, continuous missing values). Conclusions: The proposed method can retrieve any number of missing values at any location using observed data with a minimal number of calculations.

SVIoT: A Secure Visual-IoT Framework for Smart Healthcare.

The advancement of the Internet of Things (IoT) has transfigured the overlay of the physical world by superimposing digital information in various sectors, including smart cities, industry, healthcare, etc. Among the various shared information, visual data are an insensible part of smart cities, especially in healthcare. As a result, visual-IoT research is gathering momentum. In visual-IoT, visual sensors, such as cameras, collect critical multimedia information about industries, healthcare, shopping, autonomous vehicles, crowd management, etc. In healthcare, patient-related data are captured and then transmitted via insecure transmission lines. The security of this data are of paramount importance. Besides the fact that visual data requires a large bandwidth, the gap between communication and computation is an additional challenge for visual IoT system development. In this paper, we present SVIoT, a Secure Visual-IoT framework, which addresses the issues of both data security and resource constraints in IoT-based healthcare. This was achieved by proposing a novel reversible data hiding (RDH) scheme based on One Dimensional Neighborhood Mean Interpolation (ODNMI). The use of ODNMI reduces the computational complexity and storage/bandwidth requirements by 50 percent. We upscaled the original image from M × N to M ± 2N, dissimilar to conventional interpolation methods, wherein images are upscaled to 2M × 2N. We made use of an innovative mechanism, Left Data Shifting (LDS), before embedding data in the cover image. Before embedding the data, we encrypted it using an AES-128 encryption algorithm to offer additional security. The use of LDS ensures better perceptual quality at a relatively high payload. We achieved an average PSNR of 43 dB for a payload of 1.5 bpp (bits per pixel). In addition, we embedded a fragile watermark in the cover image to ensure authentication of the received content.

Physics super-resolution of thermal environment maps in urban area by deep neural network with attention

Real-time micrometeorology predictions are important for heat-stroke risk reduction and safe drone logistics in urban streets. However, the extensive calculation cost hinders operational real-time predictions of micrometeorology. In order to solve this, it is proposed to use a super-resolution simulation system that can provide high-resolution (HR) predictions at the low cost for low-resolution (LR) simulations. The system utilizes the super-resolution that converts LR maps into HR ones by means of machine-learning technologies. The super-resolution technology has been advancing rapidly and its neural networks can now learn physics to some extent. It has, however, the room for improvement. We propose the Pixel Attention Super-Resolution Network (PA-SRN), which gives attention weights to features and space at the same time. The training and test data for the deep neural networks were generated from the building-resolving urban micrometeorology simulations with a multi-scale atmosphere-ocean coupled model named the Multi-Scale Simulator for the Geoenvironment (MSSG). The proposed PA-SRN model has improved learning accuracy compared to a conventional algebraic interpolation and other existing neural network models. In addition, the analysis of attention weight distributions has clarified how the proposed neural network learns the physics in the micrometeorology data.

A Deep Two-State Gated Recurrent Unit for Particulate Matter (PM2.5) Concentration Forecasting

Air pollution is a significant problem in modern societies since it has a serious impact on human health and the environment. Particulate Matter (PM2.5) is a type of air pollution that contains of interrupted elements with a diameter less than or equal to 2.5 m. For risk assessment and epidemiological investigations, a better knowledge of the spatiotemporal variation of PM2.5 concentration in a constant space-time area is essential. Conventional spatiotemporal interpolation approaches commonly relying on robust presumption by limiting interpolation algorithms to those with explicit and basic mathematical expression, ignoring a plethora of hidden but crucial manipulating aspects. Many advanced deep learning approaches have been proposed to forecast Particulate Matter (PM2.5). Recurrent neural network (RNN) is one of the popular deep learning architectures which is widely employed in PM2.5 concentration forecasting. In this research, we proposed a Two-State Gated Recurrent Unit (TS-GRU) for monitoring and estimating the PM2.5 concentration forecasting system. The proposed algorithm is capable of considering both spatial and temporal hidden affecting elements spontaneously. We tested our model using data from daily PM2.5 dimensions taken in the contactual southeast area of the United States in 2009. In the studies, three evaluation matrices were utilized to compare the overall performance of each algorithm: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The experimental results revealed that our proposed TS-GRU model outperformed compared to the other deep learning approaches in terms of forecasting performance.

Three-Dimensional Video Super-Resolution Reconstruction Scheme Based on Histogram Matching and Recursive Bayesian Algorithms

Multimedia Super-Resolution (SR) reconstruction is an essential and mandatory process for different visualization functions. Recently, several schemes have been suggested for single- and multi-image SR reconstruction. This work presents an effective SR reconstruction scheme for visual quality and resolution enhancement of 3D Video (3DV) sequences. The idea behind the proposed 3DV SR reconstruction scheme is the utilization of a recursive Bayesian algorithm for improving the resolution of the degraded 3DV sequences with down-sampling, blurring, and noise effects. In addition, a significant stage of histogram matching based on a visual image with a better-distributed histogram is employed. The main aim of employing the histogram matching stage for enhancing the 3DV sequence is to introduce a dynamic range modification of each 3DV frame. Hence, it presents a 3DV sequence with an enhanced distribution of intensities. This improves the whole performance efficiency of the suggested scheme. The performance of the proposed SR reconstruction scheme is compared with that of the conventional bicubic interpolation scheme. Comparisons with recent and related SR reconstruction schemes are also introduced. Simulation results reveal that the proposed scheme achieves superior outcomes in terms of Structural Similarity (SSIM) index, local contrast, average gradient, Mean Square Error (MSE), edge intensity, entropy, and Peak Signal-to-Noise Ratio (PSNR) of the resulting 3DV frames.

Self-Supervised Deep Learning to Reconstruct Seismic Data With Consecutively Missing Traces

Seismic data processing requires careful interpolation or reconstruction to restore the regularly or irregularly missing traces. In practice, seismic data with consecutively missing traces are quite common, which will lead to a great challenge for conventional interpolation or reconstruction methods. To effectively reconstruct the successively blank traces in seismic data, we proposed a self-supervised deep learning approach, with which the convolutional neural network is trained in a supervised manner with pseudolabels obtained from unlabeled observed data. The pseudolabels are automatically generated by randomly masking the observed data to simulate the consecutively missing scenario. We train a nested U-Net structure (UNet++) with a hybrid loss function so that the local and global structural information can be captured to ensure the quality of reconstruction. A two-step reconstruction workflow is designed to recover the missing recordings with respect to both the receivers and sources. Synthetic and field data examples demonstrate that the proposed self-supervised learning can effectively reconstruct the corrupted seismic data.

Deep learning super-resolution magnetic resonance spectroscopic imaging of brain metabolism and mutant isocitrate dehydrogenase glioma.

BackgroundMagnetic resonance spectroscopic imaging (MRSI) can be used in glioma patients to map the metabolic alterations associated with IDH1,2 mutations that are central criteria for glioma diagnosis. The aim of this study was to achieve super-resolution (SR) MRSI using deep learning to image tumor metabolism in patients with mutant IDH glioma.MethodsWe developed a deep learning method based on generative adversarial network (GAN) using Unet as generator network to upsample MRSI by a factor of 4. Neural networks were trained on simulated metabolic images from 75 glioma patients. The performance of deep neuronal networks was evaluated on MRSI data measured in 20 glioma patients and 10 healthy controls at 3T with a whole-brain 3D MRSI protocol optimized for detection of d-2-hydroxyglutarate (2HG). To further enhance structural details of metabolic maps we used prior information from high-resolution anatomical MR imaging. SR MRSI was compared to ground truth by Mann–Whitney U-test of peak signal-to-noise ratio (PSNR), structure similarity index measure (SSIM), feature-based similarity index measure (FSIM), and mean opinion score (MOS).ResultsDeep learning SR improved PSNR by 17%, SSIM by 5%, FSIM by 7%, and MOS by 30% compared to conventional interpolation methods. In mutant IDH glioma patients proposed method provided the highest resolution for 2HG maps to clearly delineate tumor margins and tumor heterogeneity.ConclusionsOur results indicate that proposed deep learning methods are effective in enhancing spatial resolution of metabolite maps. Patient results suggest that this may have great clinical potential for image guided precision oncology therapy.

Evaluating the Efficacy of Small Face Recognition by Convolutional Neural Networks with Interpolation Based on Auto-adjusted Parameters and Transfer Learning

ABSTRACT In this work, we propose a new approach for face recognition using low-resolution images. By cleverly combining conventional interpolation methods with the state-of-the-art classification approach, i.e. convolutional neural network, we introduce a new approach to efficiently leverage low-resolution images in classification task, especially in face recognition. Besides, we also do experiments on some recent popular methods, our approach outperforms some of them. Additionally, we propose a specific transfer learning strategy based on the preexisting well-known concept dedicated to low-resolution transfer learning. It boosts performance and reduces training time significantly. We also investigate on scalability by applying Bayesian optimization for hyper-parameter search. Therefore, our approach is able to be widely applied in many kinds of datasets and low-resolution classification tasks due to automatically seeking optimal hyper-parameters, which makes our method competitive to others.

NIMG-16. DEEP LEARNING SUPER-RESOLUTION MR SPECTROSCOPIC IMAGING TO MAP TUMOR METABOLISM IN MUTANT IDH GLIOMA PATIENTS

Abstract Metabolic imaging can map spatially abnormal molecular pathways with higher specificity for cancer compared to anatomical imaging. However, acquiring high resolution metabolic maps similar to anatomical MRI is challenging in patients due to low metabolite concentrations, and alternative approaches that increase resolution by post-acquisition image processing can mitigate this limitation. We developed deep learning super-resolution MR spectroscopic imaging (MRSI) to map tumor metabolism in patients with mutant IDH glioma. We used a generative adversarial network (GAN) architecture comprised of a UNet neural network as the generator network and a discriminator network for adversarial training. For initial training we simulated a large data set of 9600 images with realistic quality for acquired MRSI to effectively train the deep learning model to upsample by a factor of four. Two types of training were performed: 1) using only the MRSI data, and 2) using MRSI and prior information from anatomical MRI to further enhance structural details. The performance of super-resolution methods was evaluated by peak SNR (PSNR), structure similarity index (SSIM), and feature similarity index (FSIM). After training on simulations, GAN was evaluated on measured MRSI metabolic maps acquired with resolution 5.2×5.2 mm2 and upsampled to 1.3×1.3 mm2. The GAN trained only on MRSI achieved PSNR = 27.94, SSIM = 0.88, FSIM = 0.89. Using prior anatomical MRI improved GAN performance to PSNR = 30.75, SSIM = 0.90, FSIM = 0.92. In the patient measured data, GAN super-resolution metabolic images provided clearer tumor margins and made apparent the tumor metabolic heterogeneity. Compared to conventional image interpolation such as bicubic or total variation, deep learning methods provided sharper edges and less blurring of structural details. Our results indicate that the proposed deep learning method is effective in enhancing the spatial resolution of metabolite maps which may better guide treatment in mutant IDH glioma patients.

Multi-point momentum interpolation correction on collocated meshes

Since the momentum interpolation method was proposed by Rhie and Chow in 1983, it has been widely used in studies of Computational Fluids Dynamics (CFD). The conventional momentum interpolation methods were designed across the edge between two neighboring cells. In this study, an alternative momentum interpolation method, called multi-point momentum interpolation correction (IC) method, is proposed. The proposed IC method is distinguished from the conventional cross-edge momentum interpolation methods by correcting and improving the edge velocity with the interpolated values of its surrounding edges. Examples including analytic, experimental and field cases demonstrated that the proposed IC method is generally capable of improving the convergence process and numerical accuracy.

Geodesic Hermite Spline Curve on Triangular Meshes

Curves on a polygonal mesh are quite useful for geometric modeling and processing such as mesh-cutting and segmentation. In this paper, an effective method for constructing C1 piecewise cubic curves on a triangular mesh M while interpolating the given mesh points is presented. The conventional Hermite interpolation method is extended such that the generated curve lies on M. For this, a geodesic vector is defined as a straightest geodesic with symmetric property on edge intersections and mesh vertices, and the related geodesic operations between points and vectors on M are defined. By combining cubic Hermite interpolation and newly devised geodesic operations, a geodesic Hermite spline curve is constructed on a triangular mesh. The method follows the basic steps of the conventional Hermite interpolation process, except that the operations between the points and vectors are replaced with the geodesic. The effectiveness of the method is demonstrated by designing several sophisticated curves on triangular meshes and applying them to various applications, such as mesh-cutting, segmentation, and simulation.