Search

This paper introduces FusionDiff, a novel unified end-to-end image fusion network based on diffusion probabilistic models. The proposed method addresses the data dependency issue present in current unified image fusion approaches. Initially, the method pre-fuses the source images after diffusion. Subsequently, a noise prediction network forecasts the noise applied to the pre-fused image. During this stage, the Spatially-Adaptive Constraint layer is employed to restrict the fusion image, thereby maximizing the preservation of respective source image features. Finally, in the reverse process, skip-sampling is employed to effectively overcome the computational time drawback of traditional diffusion probabilistic models while maintaining high-quality image generation. Compared to other diffusion model-based image fusion methods, our approach stands out for its structural simplicity and ease of training, serving as a unified image fusion method. Furthermore, compared to other unified image fusion methods, our proposed network fully leverages the generalization of the diffusion probabilistic model, achieving adaptive feature extraction by constraining the inverse process of source images. Qualitative and quantitative experimental results across four classic image fusion tasks demonstrate the superiority of our method over state-of-the-art unified image fusion methods in recent years. • We propose a unified image fusion method based on the Diffusion Probabilistic Models. • We address the issue of data dependency, requiring training on only one multi-focus dataset to be applicable to other image fusion tasks. • We designed a spatially-adaptive constraint layer to impose conditional constraints on the diffusion model. [ABSTRACT FROM AUTHOR]

• WDGPA is a heterogeneous molecular network based solution for genome-phenome association prediction. • An adaptive weighting strategy is introduced on intra/inter-association and attribution matrices to differentially fuse them. • The representations of molecules can be obtained by performing layerby-layer nonlinear factorization on intra/inter-association and attribution matrices in a collaborative manner. • Genome-phenome associations can be predicted by using representational vectors of genes and phenotypes. • WDGPA can achieve better performance and accurately predict potential gene-phenotype associations in real data. Genome-phenome association (GPA) prediction can promote the understanding of biological mechanisms about complex pathology of phenotypes (i.e., traits and diseases). Traditional heterogeneous network-based GPA approaches overwhelmingly need to project heterogeneous data toward homogeneous network for data fusion and prediction, such projections result in the loss of heterogeneous network structure information. Matrix factorization based data fusion can avoid such projection by integrating multi-type data in a coherent way, but they typically perform linear factorization and cannot mine the nonlinear relationships between molecules, which compromise the accuracy of GPA analysis. Furthermore, most of them can not selectively synergy network topology and node attribution information in a principle way. In this paper, we propose a weighted deep matrix factorization based solution (WDGPA) to predict GPAs by selectively and differentially fusing heterogeneous molecular network and diverse attributes of nodes. WDGPA firstly assigns weights to inter/intra-relational data matrices and attribute data matrices, and performs deep matrix factorization on these matrices of heterogeneous network in a cooperative manner to obtain the nonlinear representations of different nodes. In addition, it performs low-rank representation learning on the attribute data with the shared nonlinear representations. In this way, both the network topology and node attributes are jointly mined to explore the representations of molecules and complex interplays between molecules and phenotypes. WDGPA then uses the representational vectors of gene and phenotype nodes to predict GPAs. Experimental results on maize and human datasets confirm that WDGPA outperforms competitive methods by a large margin under different evaluation protocols. [ABSTRACT FROM AUTHOR]

6-DoF grasp estimation methods based on point clouds have long been a challenge in robotics due to the limitations of single data input, which hinder the robot's perception of real-world scenarios, thus reducing its robustness. In this work, we propose a 6-DoF grasp pose estimation method based on RGB-D data, which leverages ResNet to extract color image features, utilizes the PointNet++ network to extract geometric information features, and employs an external attention mechanism to fuse both features. Our method is an end-to-end design, and we validate its performance through benchmark tests on a large-scale dataset and evaluations in a simulated robot environment. Our method outperforms previous state-of-the-art methods on public datasets, achieving 47.75mAP and 40.08mAP for seen and unseen objects, respectively. We also test our grasp pose estimation method on multiple objects in a simulated robot environment, demonstrating that our approach exhibits higher grasp accuracy and robustness than previous methods. [ABSTRACT FROM AUTHOR]

• An experimental variational mode decomposition (VMD)-based data-fusion displacement measurement framework is proposed. • Vision, acceleration, and fusion methods are compared. • A four-story RC structure on a 25-ton table is detailed. Dynamic displacement is a crucial parameter in structural health monitoring (SHM) for assessing the safety, dependability, and suitability of structures under various types of excitations. Computer vision-based methods for dynamic displacement estimation have attracted much interest owing to their cost-effectiveness and convenience. However, these methods are limited by their low sampling rates and high data sensitivity. To compensate for these limitations, methods for combining data obtained from other sensors have been proposed. In this study, an experimental data-fusion framework for displacement estimation based on variational mode decomposition (VMD) was developed to leverage the advantages of vision- and acceleration-based measurements. The measurements were decomposed into ensembles of modes and recomposed to reconstruct the displacement with a higher accuracy and over a wider frequency range. An optimal mode recomposition method was proposed to achieve optimal mode combinations. Furthermore, this study introduced an improved vision-based displacement measurement method and a VMD-based indirect acceleration measurement method. The proposed framework was validated through four-story RC structure tests, which demonstrated that the method could enhance the accuracy of displacement estimation and extend the feasible frequency range compared with single-source displacement measurements. The method provides a promising solution for more effective health monitoring of modern structures subjected to a wide variety of dynamic loads. [ABSTRACT FROM AUTHOR]

• Individual response to drugs is crucial for personalized medicine. • We developed DREMO for predicting drug response in individuals. • 3 cell line similarity networks and 2 drug similarity networks were constructed. • The similarity information networks were fused with low-dimensional feature vectors. • DREMO addresses the "small n, large p" issue to a certain extent. Predicting the response of each individual patient to a drug is a key issue assailing personalized medicine. Our study predicted drug response based on the fusion of multiomics data with low-dimensional feature vector representation on a multilayer network model. We named this new method DREMO (D rug R esponse pr E diction based on M ulti O mics data fusion). DREMO fuses similarities between cell lines and similarities between drugs, thereby improving the ability to predict the response of cancer cell lines to therapeutic agents. First, a multilayer similarity network related to cell lines and drugs was constructed based on gene expression profiles, somatic mutation, copy number variation (CNV), drug chemical structures, and drug targets. Next, low-dimensional feature vector representation was used to fuse the biological information in the multilayer network. Then, a machine learning model was applied to predict new drug-cell line associations. Finally, our results were validated using the well-established GDSC/CCLE databases, literature, and the functional pathway database. Furthermore, a comparison was made between DREMO and other methods. Results of the comparison showed that DREMO improves predictive capabilities significantly. [ABSTRACT FROM AUTHOR]

• We proposed a weighted matrix factorization based data fusion approach to predict associations between lncRNAs and diseases. • Our approach can selectively integrate multiple inter-relational and intra-relational data sources. • Our approach can explore and exploit the intrinsic and shared structure of heterogeneous data sources. • Our approach is readily available for various link prediction problems. Influx evidences show that red long non-coding RNAs (lncRNAs) play important roles in various critical biological processes, and they afffect the development and progression of various human diseases. Therefore, it is necessary to precisely identify the lncRNA-disease associations. The identification precision can be improved by developing data integrative models. However, current models mainly need to project heterogeneous data onto the homologous networks, and then merge these networks into a composite one for integrative prediction. We recognize that this projection overrides the individual structure of the heterogeneous data, and the combination is impacted by noisy networks. As a result, the performance is compromised. Given that, we introduce a weighted matrix factorization model on multi-relational data to predict LncRNA-disease associations (WMFLDA). WMFLDA firstly uses a heterogeneous network to capture the inter(intra)-associations between different types of nodes (including genes, lncRNAs, and Disease Ontology terms). Then, it presets weights to these inter-association and intra-association matrices of the network, and cooperatively decomposes these matrices into low-rank ones to explore the underlying relationships between nodes. Next, it jointly optimizes the low-rank matrices and the weights. After that, WMFLDA approximates the lncRNA-disease association matrix using the optimized matrices and weights, and thus to achieve the prediction. WMFLDA obtains a much better performance than related data integrative solutions across different experiment settings and evaluation metrics. It can not only respect the intrinsic structures of individual data sources, but can also fuse them with selection. [ABSTRACT FROM AUTHOR]

[Display omitted] Rotating machines are frequently used in industrial applications. However, due to their severity, mechanical failures such as rotor imbalance, shaft imbalance, pulley imbalance, structural breakage, and bearing imbalance can lead to unplanned shutdowns. While vibration analysis-based condition monitoring techniques can detect and diagnose many early errors, certain mechanical faults have associated vibration characteristics that make it difficult to identify and distinguish these faults. To address this issue, this paper proposes a method based on data fusion for vibrational and electrical signatures to achieve new fused signatures for healthy and different faulty cases. The weighted decision fusion method generates the fused decision by weighting and combining the output of multiple sensors. Conventional vibration evaluation parameters diagnose unbalance, pulley misalignment, belt damage, and combined faults. However, these parameters have more dimensions and correlated features for some faulty cases, such as unbalance and misalignment. Therefore, the Principal Component Analysis (PCA) was applied to reduce the dimensionality of evaluating parameters and preserve almost all data variation. The PCA produces uncorrelated Principal Components (PCs) for each case. A backpropagation neural network (BPNN) was constructed to construct an integrated fault diagnosis framework. The first and second PC was inserted as input parameters in the training set of BPNN. It was observed that BPNN achieves 2.1762 × 10 - 10 Mean Squared Error (MSE) demonstrates superior data fusion solutions and PCA in the condition monitoring of rotating machines. Overall, this study proposes an effective method for diagnosing mechanical faults in rotating machines, which can improve reliability and reduce downtime in industrial applications. [ABSTRACT FROM AUTHOR]

At the heart of structural engineering research is the use of data obtained from physical structures such as bridges, viaducts and buildings. These data can represent how the structure responds to various stimuli over time when in operation. Many models have been proposed in literature to represent such data, such as linear statistical models. Based upon these models, the health of the structure is reasoned about, e.g. through damage indices, changes in likelihood and statistical parameter estimates. On the other hand, physics-based models are typically used when designing structures to predict how the structure will respond to operational stimuli. These models represent how the structure responds to stimuli under idealised conditions. What remains unclear in the literature is how to combine the observed data with information from the idealised physics-based model into a model that describes the responses of the operational structure. This paper introduces a new approach which fuses together observed data from a physical structure during operation and information from a mathematical model. The observed data are combined with data simulated from the physics-based model using a multi-output Gaussian process formulation. The novelty of this method is how the information from observed data and the physics-based model is balanced to obtain a representative model of the structures response to stimuli. We present our method using data obtained from a fibre-optic sensor network installed on experimental railway sleepers. The curvature of the sleeper at sensor and also non-sensor locations is modelled, guided by the mathematical representation. We discuss how this approach can be used to reason about changes in the structures behaviour over time using simulations and experimental data. The results show that the methodology can accurately detect such changes. They also indicate that the methodology can infer information about changes in the parameters within the physics-based model, including those governing components of the structure not measured directly by sensors such as the ballast foundation. • Probabilistic models can be used to balance information obtained from physics-based models and observed data. • Balancing is implemented by maximizing the predictive performance of observed test data-points. • Physics-based models can improve the inference of the structural response in areas of the domain which are unmeasured. • The proposed information synthesis can infer model parameters that govern structural components that cannot be measured. • Experimental application of the proposed methodology, detecting underlying changes in model parameters resulting from damage. [ABSTRACT FROM AUTHOR]

• There have been very few researches on unscented Kalman filter with unknown input. • A novel UKF-UI is proposed for recursive state-input-system identification of nonlinear systems. • The proposed UKF is derived analogously to the procedures of the conventional UKF. • Data fusion of partially measured accelerations and displacements is used to prevent the drifts in identification. • Such a presented analytical solution of UKF-UI is not available in previous literatures. The unscented Kalman filter (UKF) has proven to be an effective approach for the identification of nonlinear systems from limited output measurements. However, the conventional UKF requires that measurements of the input excitations are available to successfully perform nonlinear system identification, which limits its application in cases where it is difficult or impractical to measure the inputs. In this paper a novel unscented Kalman filter with unknown input (UKF-UI) is proposed for the simultaneous identification of nonlinear structural systems and external excitations. Based on the estimation-based procedures of the conventional UKF, the analytical recursive solutions of the proposed UKF-UI are derived in an analogous fashion resulting in a recursive nonlinear least-squares problem for the unknown input. Moreover, data fusion of partially measured acceleration and displacement responses is used to alleviate the drifts typically observed in the estimated inputs and displacements. Numerical and experimental validation examples are used to demonstrate the effectiveness of the proposed UKF-UI algorithm for the simultaneous identification of nonlinear parameters and unknown external excitations using data fusion of partially measured system responses. [ABSTRACT FROM AUTHOR]

• We use context fusion and subspace constraint to improve correlation filter tracking. • A model updating strategy for accelerating the tracker is proposed. • Ablation study on each component of the proposed method is analyzed in detail. • Extensive experiments verify the effectiveness of the proposed tracker. In the trackers based on correlation filter, the limitation of tracking samples and the efficiency of the solution affect the performance of the trackers. Based on the Discriminative Scale Space Tracking(DSST), we proposed a new framework to improve correlation filter tracking via context fusion and subspace constraint. First, the target and background are jointly modeled to enhance the ability to distinguish between target and background. Second, to achieve higher robustness, subspace constraint is applied to DSST, and it can be solved efficiently by subspace-based alternating direction method of multipliers(SADMM). Finally, through the model update strategy, high-quality samples are selected to update. The proposed algorithm is validated on the OTB2013, OTB2015 and TC128. Experimental results show that compared with fast DSST(fDSST), success rate and accuracy of the proposed method are increased by 4.2% and 6.8% respectively. Compared with the state-of-the-art trackers, the proposed method achieves a competitive result with real-time speed. [ABSTRACT FROM AUTHOR]

Lassa virus (LASV) is a notorious human pathogen in West Africa. Its class I trimeric spike complex displays a distinct architecture, and its cell entry mechanism involves unique attributes not shared by other related viruses. We determined the crystal structure of the GP2 fusion glycoprotein from the spike complex of LASV (GP2 LASV) in its post-fusion conformation. GP2 LASV adopts a canonical helical bundle configuration similarly to other viruses in its family. The core packing of GP2 LASV , however, is more organized compared to GP2 from other viruses reducing the formation of internal hydrophobic cavities. We demonstrate a link between the formation of such unfavorable hydrophobic cavities and the efficiencies of membrane fusion and cell entry. Our study suggests that LASV has evolved a more efficient membrane fusogen compared to other viruses from its family by optimizing the post-fusion configuration of its GP2 module. Unlabelled Image • GP2 of Lassa virus adopts a canonical six helical bundle conformation. • Core packing at the membrane proximal region of GP2 varies between viruses. • Unfavorable hydrophobic cavities influence membrane fusion efficiencies. • Lassa virus evolved an optimized fusogen compared to viral family members. [ABSTRACT FROM AUTHOR]

Highlights • A novel saliency prediction method with optimal scheme of color components is proposed. • A light-weight deep model with CNN is presented to learn the optimal scheme of color components. • A new fusion method of topological feature is proposed via calculating Mahalanobis distance. Abstract A new saliency prediction method via extracting topological feature and calculating Mahalanobis distance on deep color components is presented in this paper. Specifically, four selectable schemes of color components are considered and a deep convolutional network is used to learn the best scheme. Then the topological feature maps of an input image are extracted on the learned color components by the analysis of connectivity and adjacency. To achieve the final saliency map, a new fusion method is proposed by calculating the Mahalanobis distance between the feature maps and their means with their covariance matrices rather than summating the feature maps linearly. The numerical and visual evaluation shows that a competitive performance compared with fourteen state-of-the-art models is achieved by the proposed method. [ABSTRACT FROM AUTHOR]

Abstract The Pearl River Delta (PRD), one of the most polluted and populous regions of China, experienced a 28% reduction in fine particulate matter (PM 2.5) concentration between 2013 (47 μg/m3) and 2015 (34 μg/m3) under a stringent national policy known as the Air Pollution Prevention and Control Action Plan (hereafter Action Plan). In this study, the health and economic benefits associated with PM 2.5 reductions in PRD during 2013–2015 were estimated using the Environmental Benefits Mapping and Analysis Program-Community Edition (BenMAP-CE) software. To create reliable gridded PM 2.5 surfaces for BenMAP-CE calculations, a data fusion tool which incorporates the accuracy of monitoring data and the spatial coverage of predictions from the Community Multiscale Air Quality (CMAQ) model has been developed. The population-weighted average PM 2.5 concentration over PRD was predicted to decline by 24%. PM 2.5 -related mortality was estimated to decrease by more than 3800 due to decreases in stroke (48%), ischemic heart disease (IHD) (35%), chronic obstructive pulmonary disease (COPD) (10%), and lung cancer (LC) (7%). A 13% reduction in PM 2.5 -related premature deaths from these four causes yielded a large economic benefit of about 1300 million US dollars. Our research suggests that the Action Plan played a major role in reducing emissions and additional measures should be implemented to further reduce PM 2.5 pollution and protect public health in the future. Graphical abstract Image 1 Highlights • An innovative, open-source data fusion tool was developed to create air quality fields for exposure characterization. • Over 3800 avoided deaths in Pearl River Delta (PRD), China were estimated between 2013 and 2015 due to PM 2.5 reduction. • Health benefits varied spatially across PRD and were limited by the flat slope of the IER model at high concentration. [ABSTRACT FROM AUTHOR]

Highlights • Use single fisheye camera to cover a hemisphere FOV of MAV. • Use original fisheye images to implement object detection. • Propose a detector that is more accurate and faster than baselines on TX2. • Fuse fisheye model, detection results, attitude and height to localize objects. Abstract In robotic systems, the fisheye camera can provide a large field of view (FOV). Usually, the traditional restoring algorithms are needed, which are computational heavy and will introduce noise into original data, since the fisheye images are distorted. In this paper, we propose a framework to detect objects from the raw fisheye images without restoration, then locate objects in the real world coordinate by fusing attitude information. A deep neural network architecture based on the MobileNet and feature pyramid structure is designed to detect targets directly on the fisheye raw images. Then, the target can be located based on the fisheye visual model and the attitude of the camera. Compared to traditional approaches, this approach has advantages in computational efficiency and accuracy. This approach is validated by experiments with a fisheye camera and an onboard computer on a micro-aerial vehicle (MAV). [ABSTRACT FROM AUTHOR]

Abstract Panoramic visual tracking has high application value in many situations, but its visual distortion is likely to cause low tracking robustness or loss of target. This paper presents a panoramic visual tracking method based on adaptive feature fusion method, the size of the trapezoidal frame is calibrated with target moving in the method, and the linear model of trapezoidal frame parameters is fitted, then according to the model trapezoidal area of target extracted, and are modified to target trapezoidal region through the affine transformation; Based on filter tracking framework, the fusion of color and shape information is as the main characteristics of the target tracking, and Bayesian fusion recursive formula is used for calculating the particle weights. The experimental results show that the proposed algorithm is better than the existing methods in tracking precision and anti occlusion, and effectively improves the robustness of panoramic visual tracking. [ABSTRACT FROM AUTHOR]

An F -essential subgroup is called a pearl if it is either elementary abelian of order p 2 or non-abelian of order p 3 . In this paper we start the investigation of fusion systems containing pearls: we determine a bound for the order of p -groups containing pearls and we classify the saturated fusion systems on p -groups containing pearls and having sectional rank at most 4. [ABSTRACT FROM AUTHOR]

Highlights • Depth estimation from light field data. • Guided filtering-based data fusion. • Fusion of stereo disparity and defocus response. • Producing rich details and clear edges in depth. In this paper, we propose guided filtering based data fusion for light field depth estimation with L 0 gradient minimization. Stereo disparity produces good depth edge, while defocus response yields good depth information in homogeneous regions. We fuse stereo disparity and defocus response from light filed data in a guided filtering framework. In the guided filtering framework, we adopt L 0 gradient minimization as the regularization term instead of penalizing linear coefficients to consider depth characteristics that have similar depth in the same object. Moreover, we utilize edge direction in stereo matching to prevent the confusion caused by occlusion. Experimental results on both synthetic and real light field datasets show that the proposed method achieves clearer edge and less error in depth than state-of-the-arts. [ABSTRACT FROM AUTHOR]

Based on convex analysis, a novel consensus algorithm of dynamical points in a CAT(0) space is developed in this paper, in which the associated communication graph uniformly contains a directed spanning tree. The proposed algorithm provides an efficient method of solving consensus problems in a general CAT(0) space, while having certain robustness against weak communication. The application of the new algorithm to the distributed fusion of phylogenetic trees is shown with demonstrative-case simulations, together with a study on the algorithm's robustness and efficiency. [ABSTRACT FROM AUTHOR]

Fast intra mode decision strategies are proposed to overcome the brute force mode decision for the coding unit (CU) in High Efficiency Video Coding (HEVC). The proposed work improves the rough-mode-decision (RMD) by initializing the candidate intra mode list using the fusion of the Hadamard-cost and the statistical-inference formed using spatial/ temporal correlations. Then an early termination is predicted using optimal stopping theory that addresses early decision for a generic class of decision problems. Subsequently, a novel RD-cost prediction model is developed for early termination that is based on the RD-cost variation in the neighboring CUs with-respect-to their co-located CUs. Experimental results demonstrate that the RMD module of HEVC and the state-of-the-art fast intra mode prediction published method are outperformed by saving up to 0.61% and 0.91% Bjøntegaard delta bit rate (BDBR) on average, respectively. [ABSTRACT FROM AUTHOR]

Sketch face recognition (SFR) has been widely and successfully applied in law enforcement, which attracts a growing number of researchers. In this paper, a face sketch-photo synthesis and recognition method is proposed. Our method has two parts: Firstly, according to the different synthesis results for different scales, a cascade sketch-photo synthesis method via dual-scale Markov Network is utilized for image synthesis; Secondly, structural information and feature information-based data fusion method has been presented for face recognition. It is inspired by the Face Recognition Cognitive Theory, which applies both structural information and feature information for recognition. The experimental results on different databases based on the proposed method, demonstrate the outperformance of our method compared with state-of-the-art methods both in synthesis and recognition processes. [ABSTRACT FROM AUTHOR]

This paper presents a recursive algorithm where a time delay is considered in the estimation of the forces applied to a structure and the corresponding system states. In particular when the measured response is not collocated with the estimated forces, essential information on the estimated forces and/or system states is contained in the response at L consecutive time steps following the time step where the estimation is performed. The main focus in this paper is on the reduction in estimation uncertainty that can be achieved by so-called smoothing, i.e. by considering a time delay in the estimation. When the calculation of the gain matrices is included in the recursive estimation, the calculation time of the algorithm largely increases with the time delay. It is shown that a prior calculation of the steady-state gain matrices allows for a significant reduction of the calculation time. The presented algorithm is first verified using numerical simulations. Next, a validation is performed using data obtained from a field test on a footbridge. [ABSTRACT FROM AUTHOR]

We develop a general CFD framework for multi-resolution simulations to target multiscale problems but also resilience in exascale simulations, where faulty processors may lead to gappy, in space-time, simulated fields. We combine approximation theory and domain decomposition together with statistical learning techniques, e.g. coKriging, to estimate boundary conditions and minimize communications by performing independent parallel runs. To demonstrate this new simulation approach, we consider two benchmark problems. First, we solve the heat equation (a) on a small number of spatial “patches” distributed across the domain, simulated by finite differences at fine resolution and (b) on the entire domain simulated at very low resolution, thus fusing multi-resolution models to obtain the final answer. Second, we simulate the flow in a lid-driven cavity in an analogous fashion, by fusing finite difference solutions obtained with fine and low resolution assuming gappy data sets. We investigate the influence of various parameters for this framework, including the correlation kernel, the size of a buffer employed in estimating boundary conditions, the coarseness of the resolution of auxiliary data, and the communication frequency across different patches in fusing the information at different resolution levels. In addition to its robustness and resilience, the new framework can be employed to generalize previous multiscale approaches involving heterogeneous discretizations or even fundamentally different flow descriptions, e.g. in continuum-atomistic simulations. [ABSTRACT FROM AUTHOR]

Moving object tracking under complex scenes remains to be a challenging problem because the appearance of a target object can be drastically changed due to several factors, such as occlusions, illumination, pose, scale change and deformation. This study proposes an adaptive multi–feature fusion strategy, in which the target appearance is modeled based on timed motion history image with HSV color histogram features and edge orientation histogram features. The variances based on the similarities between the candidate patches and the target templates are used for adaptively adjusting the weight of each feature. Double templates matching, including online and offline template matching, is adopted to locate the target object in the next frame. Experimental evaluations on challenging sequences demonstrate the accuracy and robustness of the proposed algorithm in comparison with several state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

Due to the noise of sensing equipment, the tire states, such as the sideslip angle and the slip ratio, cannot be accurately observed under the conditions with small acceleration, which results in the inapplicability of the time domain data based tire-road friction coefficient (TRFC) estimation method. In order to overcome this shortcoming, frequency domain data fusion is proposed to estimate the TRFC based on the natural frequencies of the steering system and the in-wheel motor driving system. Firstly, a relationship between TRFC and the steering system natural frequency is deduced by investigating its frequency response function (FRF). Then the lateral TRFC is determined by the steering natural frequency which is only identified using the information of the assist motor current and the steering speed of the column. With spectral comparison between the steering and driving systems, the data fusion is carried out to get a comprehensive TRFC result, using the different frequency information of the longitudinal and lateral value. Finally, simulations and experiments on different road surfaces validated the correctness of the steering system FRF and the effectiveness of the proposed approach. [ABSTRACT FROM AUTHOR]

Object co-segmentation aims to simultaneously segment common regions of interest from multiple images. It is of great importance to image classification, object recognition and image retrieval. One way to extract similar objects shared by multiple images is to construct a correlation function between image regions. In this paper, object co-segmentation is addressed based on weakly supervised data fusion. First, we integrate the image boundary information into weakly supervised clustering by adopting an efficient image segmentation algorithm with proved convergence. Feature learning as well as clustering are also incorporated into the proposed algorithm to establish a unified framework so that an optimal feature subspace with clustering-oriented methods is provided. Second, the shared object from multiple images is regarded as the procedure to search objects from heterogeneous data sources, which is formalized as data fusion problems. Using data fusion techniques, we present a novel method to evaluate the similarity between images, which facilitates the use of similar objects from multiple images. Finally, the two proposed object segmentation and co-segmentation algorithms are verified through publicly available datasets MSRA1000 and iCoseg. Experiments demonstrate that both algorithms are capable to achieve superior or comparable performance over the compared state-of-the-art segmentation methods in all tested datasets. [ABSTRACT FROM AUTHOR]

K value is an important freshness indicator of meat. This study investigated the integration of spectral and textural data for enhancing the hyperspectral prediction ability of K value in pork meat. In this study, six feature wavebands (407, 481, 555, 578, 633, and 973 nm) were identified by successive projections algorithm (SPA). Meanwhile, the texture data of the grayscale images at the feature wavebands were extracted by gray level co-occurrence matrix (GLCM). The spectral and textural data were integrated by feature level fusion and the partial least square regression (PLSR) model built based on data fusion yielded excellent results, an improvement of at least 17.5% was obtained in model performance compared to those when either spectral data or textural data were used alone, indicating that data fusion is an effective way to enhance hyperspectral imaging ability for the determination of K values for freshness evaluation in pork meat. [ABSTRACT FROM AUTHOR]

Video based action recognition is one of the important and challenging problems in computer vision research. Bag of visual words model (BoVW) with local features has been very popular for a long time and obtained the state-of-the-art performance on several realistic datasets, such as the HMDB51, UCF50, and UCF101. BoVW is a general pipeline to construct a global representation from local features, which is mainly composed of five steps; (i) feature extraction, (ii) feature pre-processing, (iii) codebook generation, (iv) feature encoding, and (v) pooling and normalization. Although many efforts have been made in each step independently in different scenarios, their effects on action recognition are still unknown. Meanwhile, video data exhibits different views of visual patterns , such as static appearance and motion dynamics. Multiple descriptors are usually extracted to represent these different views. Fusing these descriptors is crucial for boosting the final performance of an action recognition system. This paper aims to provide a comprehensive study of all steps in BoVW and different fusion methods, and uncover some good practices to produce a state-of-the-art action recognition system. Specifically, we explore two kinds of local features, ten kinds of encoding methods, eight kinds of pooling and normalization strategies, and three kinds of fusion methods. We conclude that every step is crucial for contributing to the final recognition rate and improper choice in one of the steps may counteract the performance improvement of other steps. Furthermore, based on our comprehensive study, we propose a simple yet effective representation, called hybrid supervector , by exploring the complementarity of different BoVW frameworks with improved dense trajectories. Using this representation, we obtain impressive results on the three challenging datasets; HMDB51 (61.9%), UCF50 (92.3%), and UCF101 (87.9%). [ABSTRACT FROM AUTHOR]

This paper presents a verification of a joint input-state estimation algorithm using data obtained from in situ experiments on a footbridge. The estimation of the input and the system states is performed in a minimum-variance unbiased way, based on a limited number of response measurements and a system model. A dynamic model of the footbridge is obtained using a detailed finite element model that is updated using a set of experimental modal characteristics. The joint input-state estimation algorithm is used for the identification of two impact, harmonic, and swept sine forces applied to the bridge deck. In addition to these forces, unknown stochastic forces, such as wind loads, are acting on the structure. These forces, as well as measurement errors, give rise to uncertainty in the estimated forces and system states. Quantification of the uncertainty requires determination of the power spectral density of the unknown stochastic excitation, which is identified from the structural response under ambient loading. The verification involves comparing the estimated forces with the actual, measured forces. Although a good overall agreement is obtained between the estimated and measured forces, modeling errors prohibit a proper distinction between multiple forces applied to the structure for the case of harmonic and swept sine excitation. [ABSTRACT FROM AUTHOR]

This work is motivated by the lack of manpower to perform vineyard spraying tasks and the exposure to hazardous pesticides during the spraying. The development of an autonomous system for vineyard spraying would reduce the amount of required labour and redirect it to performing tasks that could increase the farm yield and agricultural profitability and economic survival. Localisation, which is accurately estimating the location of the robot, is a fundamental problem in the field of autonomous mobile robotics. In order to allow basic autonomous navigation of a field robot, a path-planning control law is necessary. This navigation algorithm requires knowledge of the accurate state of the robot at every instance (i.e. position, orientation, linear and angular velocity). Many methods for low cost sensors and state estimation were introduced over the years and each method is based on some assumptions that not always hold in the real field robot case. For example, many state estimation algorithms assume Gaussian noise of the sensors reading. This assumption is not always valid when dealing with GPS, or taking measures in a short time window. Hence it is required to develop an accurate state estimation algorithm that will be based on as many sensors as possible, and will use the advantage of each sensor in an optimal way. Therefore, a new data fusion algorithm is proposed for navigation, that optimally fused the localisation data from various sensors. This paper begins with the development of a kinematic model of the robot that is used for model-based state estimation (filtering). How to filter a noisy raw sensor data and the fusion of data from all sensors (DGPS, IMU and vision) are explored. A novel vineyard sprayer, and its new kinematic structure, is introduced. The methodology for designing a high precision localisation system for sensors data fusion, utilising a likelihood ratio test as a decision-making technique for choosing the most probable state estimation. Each sensor was pre-filtered according to its noise distribution. The localisation algorithm was validated using simulation of the robotic platform and using visual odometry based on real field video data. [ABSTRACT FROM AUTHOR]

The study of visible and near-infrared hyperspectral imaging (400–1000 nm) in tandem with data fusion technique was conducted to predict sensory quality index scores (QIS) of grass carp fish fillet for the first time. Five characteristic wavelength variables were selected by successive projections algorithm (SPA) and 13 textural feature variables were also extracted by grey-level gradient cooccurrence matrix (GLGCM) method. Least squares-support vector machine (LS-SVM) was used to build calibration models for predicting QIS estimated by traditional quality index method (QIM) based on full spectra, optimal spectra, image texture parameters and their combined data. The LS-SVM model established by data fusion of optimal spectra and texture data showed the best prediction performance with residual predictive deviation (RPD) of 4.23, coefficient of determination ( R P 2 ) of 0.944 and root mean square error in prediction (RMSEP) of 0.703. Image processing algorithm was then developed to transfer the best LS-SVM model to each pixel for visualizing the spatial distribution of QIS. The results showed that integration of hyperspectral imaging and data fusion coupled with LS-SVM analysis provides a successful quantitative ability for predicting and visualizing the spatial distribution of QIS in grass carp fish muscle. [ABSTRACT FROM AUTHOR]

This paper studies the feasibility of combining hyperspectral imaging (HSI) and electronic nose (E-nose) techniques to improve the prediction performance of moisture content (MC) in frozen-thawed pork. In this experiment, the data of 240 pork samples under different freeze-thaw cycles (0–5 times) were collected using HSI and E-nose technology. Afterward, the spectra and image (color and texture) information were extracted from the HSI sensor, while the odor information was extracted from the E-nose sensor. In this setting, an improved decision fusion method was proposed to fuse this information, thus detecting the MC in pork. Based on single information and its fusion at different fusion levels (pixel, feature, traditional decision fusion, and improved decision fusion), partial least squares regression (PLSR) prediction models were established and compared. The results show that the model of improved decision fusion based on HSI and E-nose fusion technology exhibits the best prediction capability with the determination coefficient of prediction set (R2p) of 0.9533 and root mean square error of prediction set (RMSEP) of 0.3869. This study demonstrates that the proposed method is an effective data fusion method, and the combination of HSI and E-nose technology improves the prediction performance of MC in frozen-thawed pork. • HSI and E-nose were combined to predict moisture content of frozen-thawed pork. • A decision fusion method was proposed to improve the prediction performance. • The performance of PLSR model was enhanced by the decision fusion method. • The optimal features were identified by MI-VIF. [ABSTRACT FROM AUTHOR]

Adjustment of the water treatment process to changes in water quality is a focus area for engineers and managers of water treatment plants. The desired and preferred capability depends on timely and quantitative knowledge of water quality monitoring in terms of total suspended solids (TSS) concentrations. This paper presents the development of a suite of nowcasting and forecasting methods by using high-resolution remote-sensing-based monitoring techniques on a daily basis. First, the integrated data fusion and mining (IDFM) technique was applied to develop a near real-time monitoring system for daily nowcasting of the TSS concentrations. Then a nonlinear autoregressive neural network with external input (NARXNET) model was selected and applied for forecasting analysis of the changes in TSS concentrations over time on a rolling basis onward using the IDFM technique. The implementation of such an integrated forecasting and nowcasting approach was assessed by a case study at Lake Mead hosting the water intake for Las Vegas, Nevada, in the water-stressed western U.S. Long-term monthly averaged results showed no simultaneous impact from forest fire events on accelerating the rise of TSS concentration. However, the results showed a probable impact of a decade of drought on increasing TSS concentration in the Colorado River Arm and Overton Arm. Results of the forecasting model highlight the reservoir water level as a significant parameter in predicting TSS in Lake Mead. In addition, the R-squared value of 0.98 and the root mean square error of 0.5 between the observed and predicted TSS values demonstrates the reliability and application potential of this remote sensing-based early warning system in terms of TSS projections at a drinking water intake. [ABSTRACT FROM AUTHOR]

Total volatile basic nitrogen (TVB-N) content is an important indicator in evaluating pork meat's freshness. This paper attempted a new strategy for measurement of TVB-N content in the pork meat by integrating two nondestructive sensing tools of hyperspectral imaging (HSI) and colorimetric sensors. First, HSI system and colorimetric sensors array system were used for data acquisition and further data processing, respectively. Herein, we proposed a novel efficient back propagation adaptive boosting (BP-AdaBoost) algorithm for data fusion and modeling, and we compared it with the classic modeling algorithm based on principal component analysis and back propagation artificial neural network (PCA-BPANN). Experiments results showed that the model based on data fusion was superior to the model based on the single sensing tool, and BP-AdaBoost has stronger capacity in solution to the complicated data fusion in contrast the PCA-BPANN. And the optimum results were achieved with the ratio of prediction to deviation ( RPD ) = 2.885, and the correction coefficient ( R ) = 0.932 in the prediction set. This work demonstrates that it has the potential to nondestructive detection of TVB-N content in pork meat by integrating HSI technique and colorimetric sensors technique combined with BP-AdaBoost nonlinear data fusion algorithm. [ABSTRACT FROM AUTHOR]

In this work, data fusion of multi-element XRF results from archaeological feature soils and regional background soils was applied to assess the complementary value of geochemistry and machine-learning on predictive modelling in archaeology. Our principal aim was to integrate multiple data sources, train learning models for classification of archaeological soils and background soils, and compare model predictions for three validation areas with current archaeological interpretation and established predictive models. This was done using three supervised machine-learning algorithms (k-nearest neighbors, support vector machines and artificial neural networks) which were trained, cross-validated and tested. The validation areas included a high archaeological potential area (n = 247 samples), the Dutch province of Zeeland (n = 261 samples) and an excavated imprint of an ancient farmhouse (n = 38 samples). The predictive models showed good overall performance and correctly classified about 95% of all test instances. According to the learning models, the first validation site has a top soil horizon that shows limited parallels with archaeological horizons used in model training, whereas features of high archaeological probability become more apparent below this horizon. This is in good correspondence with geochemical depth profiles and current archaeological interpretation. As for the second validation site, the models highlighted several archaeological hotspots that to some extend spatially coincide with areas of high archaeological potential as indicated by established predictive modelling. Reversely, the classifiers attributed high archaeological potential status to the most southern region of Zeeland, thereby complementing established modelling results. For the third validation site none of the instances were correctly classified and these results clearly show the limitations of geochemical predictive modelling of significantly different soil types (fine-to-coarse sands) compared to the training set (clayey sands). Present proof-of-concept study shows that modelling of multiple-source geochemical soil data using machine-learning algorithms can be successfully accomplished and that model predictions nicely complement current interpretation and/or established archeological predictive modelling of areas of archaeological interest. Limitations of our approach were found to reside in lithological differences between sites used for model training and prediction sites. [ABSTRACT FROM AUTHOR]

Grand global challenges of our time, such as food security and soil security, cannot be met without up-to-date, high-quality, high-resolution, spatiotemporal, and continuous soil and environmental data that characterize soil ecosystems. At local and regional scales, accurate and precise soil assessment is critical for management, soil health, and sustainability. This article presents integration pathways fusing lab- and field-based soil measurements, proximal and remote sensor data, environmental covariates, and/or methods within the framework of the Meta Soil Model which is poised to extend contemporary soil applications. The STEP-AWBH model allows to quantify soil-environmental covariates (S: soil, T: topography, E: ecology, P: parent material, A: atmosphere, W: water, B: biota, H: human factors) of which numerous can be sensed. We present an in-depth overview of proximal and remote sensor technologies that are used in the realm of digital soil assessment. Specific attention is given to the fusion process of (1) proximal, (2) proximal/remote, and (3) remote sensors to directly sense or predict soil properties. We highlight the promises and perils of sensor-derived proxies that allow inferences on soil properties and their change. From our review it is evident that there is no such single sensor or method that fits all soil applications. In many studies the fusion/integration of data and methods enhance the capabilities to assess specific soil properties. We critically contrast the benefits and constraints of proximal and remote sensing, fusion of soil-environmental data, and integration pathways to mash data and methods into complex soil assessments. [ABSTRACT FROM AUTHOR]

Lake eutrophication is a critical issue in the interplay of water supply, environmental management, and ecosystem conservation. Integrated sensing, monitoring, and modeling for a holistic lake water quality assessment with respect to multiple constituents is in acute need. The aim of this paper is to develop an integrated algorithm for data fusion and mining of satellite remote sensing images to generate daily estimates of some water quality parameters of interest, such as chlorophyll a concentrations and water transparency, to be applied for the assessment of the hypertrophic Albufera de Valencia. The Albufera de Valencia is the largest freshwater lake in Spain, which can often present values of chlorophyll a concentration over 200 mg m −3 and values of transparency (Secchi Disk, SD) as low as 20 cm. Remote sensing data from Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat Thematic Mapper (TM) and Enhance Thematic Mapper (ETM+) images were fused to carry out an integrative near-real time water quality assessment on a daily basis. Landsat images are useful to study the spatial variability of the water quality parameters, due to its spatial resolution of 30 m, in comparison to the low spatial resolution (250/500 m) of MODIS. While Landsat offers a high spatial resolution, the low temporal resolution of 16 days is a significant drawback to achieve a near real-time monitoring system. This gap may be bridged by using MODIS images that have a high temporal resolution of 1 day, in spite of its low spatial resolution. Synthetic Landsat images were fused for dates with no Landsat overpass over the study area. Finally, with a suite of ground truth data, a few genetic programming (GP) models were derived to estimate the water quality using the fused surface reflectance data as inputs. The GP model for chlorophyll a estimation yielded a R 2 of 0.94, with a Root Mean Square Error (RMSE) = 8 mg m −3 , and the GP model for water transparency estimation using Secchi disk showed a R 2 of 0.89, with an RMSE = 4 cm. With this effort, the spatiotemporal variations of water transparency and chlorophyll a concentrations may be assessed simultaneously on a daily basis throughout the lake for environmental management. [ABSTRACT FROM AUTHOR]

An improved classification of Oolong tea with different varieties is presented combining two novel artificial sensing tools (i.e. gustatory sensors and olfactory sensors). Herein, the gustatory sensors system was developed by using four electrodes (gold, copper, platinum and glassy carbon) in a standard three-electrode configuration, and the olfactory sensors system was developed based on a colorimetric sensors array. Initially, the data obtained from the two sensor systems was analyzed separately. Then, the potential of the combination of two sensors systems for classification was investigated. Principal component analysis (PCA) and linear discriminant analysis (LDA), as two classification tools, were used for data classification. The results show that the discrimination capability of the combined system is superior to that obtained with the two sensors systems separately, and eventually LDA achieved 100% classification rate by cross-validation. This work indicates that the combination of gustatory sensors system and olfactory sensors system can be a useful tool for the classification of Oolong tea with different varieties. [ABSTRACT FROM AUTHOR]

This study investigated the potential of visible and near infrared (Vis/NIR) hyperspectral imaging (HSI) to differentiate between free-range and broiler chicken meats. 120 hyperspectral images of chicken fillets were acquired and then calibrated for reflectance. Spectral data were extracted from the region of interest (ROI), followed by multiple scatter correction (MSC) to reduce the noise. Successive projection algorithm (SPA) was used to select optimal wavelengths from the full spectra. On the other hand, principal component analysis (PCA) was applied to select optimum characteristic images, and the first two principal component (PC) images were selected because PC1 and PC2 explained over 95% of variances of all spectra. Then, gray-level gradient co-occurrence matrix (GLGCM) was implemented on PC1 and PC2 images to extract 30 textural variables in total. Based on data fusion, classification models were established, in which the radial basis function-support vector machine (RBF-SVM) model gave the best results with high correct classification rate (CCR) of 93.33% for the prediction samples, demonstrating that combining spectra with texture data were effective for differentiating between free-range and broiler chicken meats. [ABSTRACT FROM AUTHOR]

The fusion rules for vertex operator algebra V L 2 A 4 are determined. [ABSTRACT FROM AUTHOR]

Let V be a vertex operator algebra. The fusion products in this paper are defined by logarithmic intertwining operators. Under this setting, we prove the pure exactness of any extension of the V -module V . Namely, if 0 → Q → ϵ P → ρ V → 0 is a short exact sequence of V -modules and W is a V -module, then 0 → Q ⊠ W → ϵ ⊠ 1 W P ⊠ W → ρ ⊠ 1 W V ⊠ W → 0 is also exact, where we view it as a sequence of g ( V ) -modules. [ABSTRACT FROM AUTHOR]

We consider the Burnside ring A ( F ) of F -stable S -sets for a saturated fusion system F defined on a p -group S . It is shown by S.P. Reeh that the monoid of F -stable sets is a free commutative monoid with canonical basis { α P } . We give an explicit formula that describes α P as an S -set. In the formula we use a combinatorial concept called broken chains which we introduce to understand inverses of modified Möbius functions. [ABSTRACT FROM AUTHOR]

The sensitive feature extraction from vibration signals is still a great challenge for effective fault classification of rolling element bearing. Current fault classification generally depends on feature pattern difference of different fault classes. This paper explores the active role of healthy pattern in fault classification and proposes a new fusion feature extraction method based on locality preserving projections (LPP). This study intends to discover the local feature pattern difference between each bearing status and the healthy condition to characterize and discriminate different bearing statuses. Specifically, the proposed fusion feature is achieved by two main steps. In the first step, a two-class model is firstly constructed for each class by using this class of signals and healthy condition signals. Then a fusion mapping is generated by mathematically combing the mappings of the LPP or its improvement for all two-class models. In the second step, the LPP is further applied to reduce the fusion mapping dimension, which is to find more sensitive low-dimensional information hidden in the high-dimensional fusion feature structure. The final achieved fusion feature can enhance the discrimination between all classes by improving the between-class scatter and within-class scatter for fault classification. Experimental results using different bearing fault types and severities under different loads show that the proposed method is well-suited and effective for bearing fault classification. [ABSTRACT FROM AUTHOR]

Multimodal fusion is becoming more common as it proves to be a powerful approach to identify complementary information from multimodal datasets. However, simulation of joint information is not straightforward. Published approaches mostly employ limited, provisional designs that often break the link between the model assumptions and the data for the sake of demonstrating properties of fusion techniques. This work introduces a new approach to synthetic data generation which allows full-compliance between data and model while still representing realistic spatiotemporal features in accordance with the current neuroimaging literature. The focus is on the simulation of joint information for the verification of stochastic linear models, particularly those used in multimodal data fusion of brain imaging data. Our first goal is to obtain a benchmark ground-truth in which estimation errors due to model mismatch are minimal or none. Then we move on to assess how estimation is affected by gradually increasing model discrepancies toward a more realistic dataset. The key aspect of our approach is that it permits complete control over the type and level of model mismatch, allowing for more educated inferences about the limitations and caveats of select stochastic linear models. As a result, impartial comparison of models is possible based on their performance in multiple different scenarios. Our proposed method uses the commonly overlooked theory of copulas to enable full control of the type and level of dependence/association between modalities, with no occurrence of spurious multimodal associations. Moreover, our approach allows for arbitrary single-modality marginal distributions for any fixed choice of dependence/association between multimodal features. Using our simulation framework, we can rigorously challenge the assumptions of several existing multimodal fusion approaches. Our study brings a new perspective to the problem of simulating multimodal data that can be used for ground-truth verification of various stochastic multimodal models available in the literature, and reveals some important aspects that are not captured or are overlooked by ad hoc simulations that lack a firm statistical motivation. [ABSTRACT FROM AUTHOR]

Multimodal neuroimaging has become a mainstay of basic and cognitive neuroscience in humans and animals, despite challenges to consider when acquiring and combining non-redundant imaging data. Multimodal data integration can yield important insights into brain processes and structures in addition to spatiotemporal resolution complementarity, including: a comprehensive physiological view on brain processes and structures, quantification, generalization and normalization, and availability of biomarkers. In this review, we discuss data acquisition and fusion in multimodal neuroimaging in the context of each of these potential merits. However, limitations – due to differences in the neuronal and structural underpinnings of each method – have to be taken into account when modeling and interpreting multimodal data using generative models. We conclude that when these challenges are adequately met, multimodal data fusion can create substantial added value for neuroscience applications making it an indispensable approach for studying the brain. [ABSTRACT FROM AUTHOR]

A fusion category of rank 4 has either four self-dual simple objects or exactly two self-dual simple objects. We study fusion categories of rank 4 with exactly two self-dual simple objects, giving nearly a complete classification of those based rings that admit pseudo-unitary categorification. More precisely, we show that if C is such a fusion category, then its Grothendieck ring K(C) must be one of seven based rings, six of which have known categorifications. In doing so, we classify all based rings associated with near-group categories of the group Z/3Z. [ABSTRACT FROM AUTHOR]

In this article we find the fusion rules of the free wreath product quantum groups Γ ˆ ≀ ⁎ S N + for any discrete group Γ . To do this we describe the spaces of intertwiners between basic corepresentations which allows us to identify the irreducible corepresentations. We then apply the knowledge of the fusion rules to prove, in most cases, several operator algebraic properties of the associated reduced C ⁎ -algebras such as simplicity and uniqueness of the trace. We also prove that the associated von Neumann algebra is a full type II 1 -factor and that the dual of Γ ˆ ≀ ⁎ S N + has the Haagerup approximation property for all finite groups Γ . [ABSTRACT FROM AUTHOR]

Fruit flies are pests of major economic importance in agriculture. Among these pests it is possible to highlight some species of genus Anastrepha , which attack a wide range of fruits, and are widely distributed in the American tropics and subtropics. Researchers seek to identify fruit flies in order to implement management and control programs as well as quarantine restrictions. However, fruit fly identification is manually performed by scarce specialists through analysis of morphological features of the mesonotum, wing, and aculeus. Our objective is to find solid knowledge that can serve as a basis for the development of a sounding automatic identification system of the Anastrepha fraterculus group, which is of high economic importance in Brazil. Wing and aculeus images datasets from three specimens have been used in this work. The experiments using a classifier multimodal fusion approach shows promising effectiveness results for identification of these fruit flies, with more than 98% classification accuracy, a remarkable result for this difficult problem. [ABSTRACT FROM AUTHOR]

Highlights: [•] Background likelihood estimation based on temporal pixel intensity analysis. [•] A new type of filtering that preserves edges in a signal. [•] Information fusion from different domains and its integration into an MRF. [•] An implementation framework capable of producing VH for highly challenging situations. [Copyright &y& Elsevier]