Optimal Data Fusion Research Articles

Multisensor data fusion of operational sea ice observations

Multisensor data fusion (MDF) is a process/technique of combining observations from multiple sensors to provide a more robust, accurate and complete description of the concerned object, environment or process. In this paper we introduce a new MDF method, multisensor optimal data fusion (MODF), to fuse different operational sea ice observations around Svalbard. The overall MODF includes regridding, univariate multisensor optimal data merging (MODM), multivariate check of consistency, and generation of new variables. For MODF of operational sea ice observations around Svalbard, the AMSR2 sea ice concentration (SIC) is firstly merged with the Norwegian Meteorological Institute ice chart. Then the daily SMOS sea ice thickness (SIT) is merged with the weekly CS2SMOS SIT to form a daily CS2SMOS SIT, which is further refined to be consistent with the SIC through consistency check. Finally sea ice volume (SIV) and its uncertainty are calculated based on the merged SIC and fused SIT. The fused products provide an improved, united, consistent and multifaceted description for the operational sea ice observations, they also provide consistent descriptions of sea ice edge and marginal ice zone. We note that uncertainties may vary during the regridding process, and therefore correct determination of the observation uncertainties is critically important for MDF. This study provides a basic framework for managing multivariate multisensor observations.

Integration of electronic nose, electronic tongue, and colorimeter in combination with chemometrics for monitoring the fermentation process of Tremella fuciformis

This study proposes an efficient method for monitoring the submerged fermentation process of Tremella fuciformis (T. fuciformis) by integrating electronic nose (e-nose), electronic tongue (e-tongue), and colorimeter sensors using a data fusion strategy. Chemometrics was employed to establish qualitative identification and quantitative prediction models. The Pearson correlation analysis was applied to extract features from the e-nose and tongue sensor arrays. The optimal sensor arrays for monitoring the submerged fermentation process of T. fuciformis were obtained, and four different data fusion methods were developed by incorporating the colorimeter data features. To achieve qualitative identification, the physicochemical data and principal component analysis (PCA) results were utilized to determine three stages of the fermentation process. The fusion signal based on full features proved to be the optimal data fusion method, exhibiting the highest accuracy across different models. Notably, random forest (RF) was shown to be the most accurate pattern recognition method in this paper. For quantitative prediction, partial least squares regression (PLSR) and support vector regression (SVR) were employed to predict the sugar content and dry cell weight during fermentation. The best respective predictive R2 values for reducing sugar, tremella polysaccharide and dry cell weight were found to be 0.965, 0.988, and 0.970. Furthermore, due to its ability to capture nonlinear data relationships, SVR had superior performance in prediction modeling than PLSR. The results demonstrated that the combination of electronic sensor fusion signals and chemometrics provided a promising method for effectively monitoring T. fuciformis fermentation.

Anomaly Detection Method for Rocket Engines Based on Convex Optimized Information Fusion.

The power system, as a core component of a launch vehicle, has a crucial impact on the reliability and safety of a rocket launch. Due to the limited measurement information inside the engine, it is often challenging to realize fast and accurate anomaly detection. For this reason, this paper introduces the rocket flight state data to expand the information source for anomaly detection. However, engine measurement and rocket flight state information have different data distribution characteristics. To find the optimal data fusion scheme for anomaly detection, a data set information fusion algorithm based on convex optimization is proposed, which solves the optimal fusion parameter using the convex quadratic programming problem and then adopts the adaptive CUSUM algorithm to realize the fast and accurate anomaly detection of engine faults. Numerical simulation tests show that the algorithm proposed in this paper has a higher detection accuracy and lower detection time than the traditional algorithm.

Structural health monitoring via a group-theoretic WSA for optimal feature selection and data fusion

In this study, two binary versions of the Water Strider Algorithm (WSA) are proposed and applied to optimal feature selection in classification problems. In the new binary versions, the formulations of WSA in continuous space are converted into binary space using group-theoretic operators (in AWSA) and sigmoid function (in BWSA). AWSA, BWSA, genetic algorithm (GA), and binary particle swarm optimization (BPSO) are selected and compared over eighteen well-known datasets from the University of California, Irvine repository. The results of AWSA indicate its satisfactory performance compared to those of other algorithms. Then, they are applied to find optimal features of a structural health monitoring classification problem using two well-known machine learning classifiers, namely k-Nearest Neighbor (KNN) and Naïve Bayes (NB) algorithms. To further improve the accuracy of the classification models, a decision-level data fusion technique is proposed based on the improved Dempster-Shafer theory. It is demonstrated that the AWSA presents superior results compared to the other algorithms and the suggested decision-level data fusion provides a reliable detection of damage.

Multi-fidelity machine learning based uncertainty quantification of progressive damage in composite laminates through optimal data fusion

Recently machine learning (ML) based approaches have gained significant attention in dealing with computationally intensive analyses such as uncertainty quantification of composite laminates. However, high-fidelity ML model construction is computationally demanding for such high-dimensional problems due to the required large amount of high-fidelity training data. We propose to address this issue effectively through multi-fidelity ML based surrogates which can use a training dataset consisting of optimally distributed high- and low-fidelity simulations. For forming multi-fidelity surrogates of progressive damage in composite laminates, we combine low-fidelity finite element analysis data obtained using Matzenmiller damage model with Hasin failure criteria and high-fidelity finite element analysis data obtained using three-dimensional continuum damage mechanics based model with P Linde’s failure criteria. It is shown that there is a significant computational advantage to using the multi-fidelity surrogate approach as compared to conventional single-fidelity surrogates. Such computational advantage through optimal data fusion without compromising accuracy becomes crucial for the subsequent data-driven uncertainty quantification and sensitivity analysis of composites involving thousands of realizations. Ply orientations come out to be the most sensitive parameters to matrix damage, fibre damage and reaction force in composite laminates. The degree of uncertainty in the output quantities depend on the input-level stochastic variations. For example, a combined stochastic variation of ±10% in material properties and ±10° in ply orientations lead to 1.85%, 16.98% and 11.24% coefficient of variation in the matrix damage, fibre damage and reaction force respectively. In general, the numerical results obtained based on the efficient data-driven approach strongly suggest that source-uncertainty of composites significantly influences the progressive damage evolution and global mechanical behaviour, leading to the realization of the importance of adopting an inclusive analysis framework considering such inevitable random variabilities.

Effective Techniques for Multimodal Data Fusion: A Comparative Analysis.

Data processing in robotics is currently challenged by the effective building of multimodal and common representations. Tremendous volumes of raw data are available and their smart management is the core concept of multimodal learning in a new paradigm for data fusion. Although several techniques for building multimodal representations have been proven successful, they have not yet been analyzed and compared in a given production setting. This paper explored three of the most common techniques, (1) the late fusion, (2) the early fusion, and (3) the sketch, and compared them in classification tasks. Our paper explored different types of data (modalities) that could be gathered by sensors serving a wide range of sensor applications. Our experiments were conducted on Amazon Reviews, MovieLens25M, and Movie-Lens1M datasets. Their outcomes allowed us to confirm that the choice of fusion technique for building multimodal representation is crucial to obtain the highest possible model performance resulting from the proper modality combination. Consequently, we designed criteria for choosing this optimal data fusion technique.

Deep Learning of GNSS Acquisition.

Signal acquisition is a crucial step in Global Navigation Satellite System (GNSS) receivers, which is typically solved by maximizing the so-called Cross-Ambiguity Function (CAF) as a hypothesis testing problem. This article proposes to use deep learning models to perform such acquisition, whereby the CAF is fed to a data-driven classifier that outputs binary class posteriors. The class posteriors are used to compute a Bayesian hypothesis test to statistically decide the presence or absence of a GNSS signal. The versatility and computational affordability of the proposed method are addressed by splitting the CAF into smaller overlapping sections, which are fed to a bank of parallel classifiers whose probabilistic results are optimally fused to provide a so-called probability ratio map from which acquisition is decided. Additionally, the article shows how noncoherent integration schemes are enabled through optimal data fusion, with the goal of increasing the resulting classifier accuracy. The article provides simulation results showing that the proposed data-driven method outperforms current CAF maximization strategies, enabling enhanced acquisition at medium-to-high carrier-to-noise density ratios.

Towards optimal deep fusion of imaging and clinical data via a model-based description of fusion quality.

Due to intrinsic differences in data formatting, data structure, and underlying semantic information, the integration of imaging data with clinical data can be non-trivial. Optimal integration requires robust data fusion, that is, the process of integrating multiple data sources to produce more useful information than captured by individual data sources. Here, we introduce the concept of fusion quality for deep learning problems involving imaging and clinical data. We first provide a general theoretical framework and numerical validation of our technique. To demonstrate real-world applicability, we then apply our technique to optimize the fusion of CT imaging and hepatic blood markers to estimate portal venous hypertension, which is linked to prognosis in patients with cirrhosis of the liver. To develop a measurement method of optimal data fusion quality deep learning problems utilizing both imaging data and clinical data. Our approach is based on modeling the fully connected layer (FCL) of a convolutional neural network (CNN) as a potential function, whose distribution takes the form of the classical Gibbs measure. The features of the FCL are then modeled as random variables governed by state functions, which are interpreted as the different data sources to be fused. The probability density of each source, relative to the probability density of the FCL, represents a quantitative measure of source-bias. To minimize this source-bias and optimize CNN performance, we implement a vector-growing encoding scheme called positional encoding, where low-dimensional clinical data are transcribed into a rich feature space that complements high-dimensional imaging features. We first provide a numerical validation of our approach based on simulated Gaussian processes. We then applied our approach to patient data, where we optimized the fusion of CT images with blood markers to predict portal venous hypertension in patients with cirrhosis of the liver. This patient study was based on a modified ResNet-152 model that incorporates both images and blood markers as input. These two data sources were processed in parallel, fused into a single FCL, and optimized based on our fusion quality framework. Numerical validation of our approach confirmed that the probability density function of a fused feature space converges to a source-specific probability density function when source data are improperly fused. Our numerical results demonstrate that this phenomenon can be quantified as a measure of fusion quality. On patient data, the fused model consisting of both imaging data and positionally encoded blood markers at the theoretically optimal fusion quality metric achieved an AUC of 0.74 and an accuracy of 0.71. This model was statistically better than the imaging-only model (AUC = 0.60; accuracy = 0.62), the blood marker-only model (AUC = 0.58; accuracy = 0.60), and a variety of purposely sub-optimized fusion models (AUC = 0.61-0.70; accuracy = 0.58-0.69). We introduced the concept of data fusion quality for multi-source deep learning problems involving both imaging and clinical data. We provided a theoretical framework, numerical validation, and real-world application in abdominal radiology. Our data suggests that CT imaging and hepatic blood markers provide complementary diagnostic information when appropriately fused.

Robust Unscented Kalman Filter-Based Decentralized Multisensor Information Fusion for INS/GNSS/CNS Integration in Hypersonic Vehicle Navigation

INS/GNSS/CNS (inertial navigation system/global navigation satellite system/celestial navigation system) integration is a favourable navigation mode for hypersonic vehicles. However, since the measurements from GNSS and CNS are easily interfered during highly dynamic maneuvers, this integration is very difficult to achieve optimal navigation solution with existing information fusion techniques. This paper proposes a new method of decentralized multi-sensor information fusion based on robust UKF (unscented Kalman filter) for INS/GNSS/CNS integration to solve the above issue. Firstly, a fault detection-based robust UKF is established for local state estimation, in which a hypothesis test is constructed via the Mahalanobis distance of innovation to detect abnormal measurements in GNSS and CNS; and subsequently, a scalar factor is determined and further introduced into the innovation covariance to decrease the Kalman gain to improve the UKF robustness against abnormal measurements. Secondly, the traditional multi-sensor optimal data fusion technique is extended to nonlinear systems by use of unscented transformation in the framework of minimum variance estimation to fuse the local state estimations from INS/GNSS and INS/CNS subsystems. The proposed information fusion method can achieve the globally optimal fusion estimation results against abnormal measurements for hypersonic vehicle navigation with INS/GNSS/CNS integration. Semi-physical simulations and comparison analysis have validated the superior performance of the proposed method.

A comprehensive comparison of data fusion approaches to multi-source precipitation observations: a case study in Sichuan province, China.

With the complex landform and climate in the Sichuan region, the need for practical and scientific research production by only utilising the rainfall data derived from ground stations or satellites has not been satisfied. To overcome this difficulty, rainfall data from 161 meteorological stations in 2016 are used in this study. According to the distribution of stations, 146 rainfall data from 161 meteorological stations in 2016 are used for inverse distance weighted interpolation, and then, linear regression, weighted regression, and Kalman filter fusion and optimal interpolation method data fusion are performed with TRMM 3B42 satellite rainfall data, respectively. Then, 15 meteorological stations evenly distributed in the study area are selected for the accuracy test. The results show that compared with the measurement at ground stations, linear regression shows the best merging effect on rainfall data derived from ground stations and satellite rainfall estimates across the daily scale: the correlation coefficient is the most significantly improved (0.2-0.7) and the reduction in root-mean-square error (RMSE) is the largest. The method is applicable for use in Sichuan Province when merging rainfall data. At the monthly scale, the rainfall data processed by using the Kalman filter present the highest accuracy (0.72-0.84). At this scale, the Kalman filter is more suitable.

Error Analysis and Modeling of GPM Dual-Frequency Precipitation Radar Near-Surface Rainfall Product

Abstract The error characterization of rainfall products of spaceborne radar is essential for better applications of radar data, such as multisource precipitation data fusion and hydrological modeling. In this study, we analyzed the error of the near-surface rainfall product of the dual-frequency precipitation radar (DPR) on the Global Precipitation Measurement Mission (GPM) and modeled it based on ground C-band dual-polarization radar (CDP) data with optimization rainfall retrieval. The comparison results show that the near-surface rainfall data were overestimated by light rain and slightly underestimated by heavy rain. The error of near-surface rainfall of the DPR was modeled as an additive model according to the comparison results. The systematic error of near-surface rainfall was in the form of a quadratic polynomial, while the systematic error of stratiform precipitation was smaller than that of convective precipitation. The random error was modeled as a Gaussian distribution centered from −1 to 0 mm h−1. The standard deviation of the Gaussian distribution of convective precipitation was 1.71 mm h−1, and the standard deviation of stratiform precipitation was 1.18 mm h−1, which is smaller than that of convective precipitation. In view of the precipitation retrieval algorithm of DPR, the error causes were analyzed from the reflectivity factor (Z) and the drop size distribution (DSD) parameters (Dm, Nw). The high accuracy of the reflectivity factor measurement results in a small systematic error. Importantly, the negative bias of Nw was very obvious when the rain type was convective precipitation, resulting in a large random error. Significance Statement This study first compares the total and different rain types of near-surface rainfall measured by DPR and ground-based radar CDP, then separates the error of DPR near-surface rainfall into systematic and random errors and analyzes the possible causes of the error. The purpose of this study is to better apply the error model to applications such as optimal data fusion and hydrological modeling, and the analysis of the error can also provide a basis for improving the spaceborne precipitation retrieval algorithm.

Throughput-Oriented Full-Duplex Cognitive Radio Network Parameter Optimization

Full-duplex cooperative spectrum sensing (FD-CSS) is an important research field in the field of spectrum sensing. In the FD-CSS network, the secondary user (SU) senses the usage status of the authorized spectrum by the primary user (PU) through the sensing channel and then reports the perceived data to the fusion center (FC) through the reporting channel. The FC makes a comprehensive judgment after summarizing the data through the fusion algorithm. In the secondary network with SU, throughput is an important index to measure the performance of the network. Taking throughput as the optimization goal, this paper theoretically deduces and verifies the optimal data fusion algorithm in cooperative spectrum sensing (CSS), the threshold of optimal energy detection, and the optimal transmission power of SU in the secondary network. The simulation results show the correctness of the results in this paper.

An optimal evidential data fusion algorithm based on the new divergence measure of basic probability assignment

An optimal evidential data fusion algorithm based on the new divergence measure of basic probability assignment

Research on Positioning Accuracy of Passive Acoustic Positioning System Based on Feature Matching in Air

Passive acoustic localization is an important interdisciplinary research branch which combines acoustics, communication and signal processing. Passive acoustic positioning system, as an important part of battlefield sensing and monitoring system, has been paid more and more attention. Due to the disturbance of battlefield and other environments and the existence of measurement errors, the positioning accuracy often cannot meet the requirements. Passive acoustic positioning system, as an important part of battlefield sensing and monitoring system, has been paid more and more attention. Due to the disturbance of battlefield and other environments and the existence of measurement errors, the positioning accuracy often cannot meet the requirements. In multi-sensor passive location, the measured data of each array element is not only related, but also compensatory and redundant, so some people use the optimal data fusion algorithm as post-processing algorithm to improve the positioning accuracy. In this paper, the positioning accuracy of ground passive acoustic positioning system is analyzed based on feature matching method, and some useful conclusions are drawn.

A novel optimal data fusion algorithm and its application for the integrated navigation system of missile

For Inertial Navigation System (INS)/Celestial Navigation System (CNS)/Global Navigation Satellite System (GNSS) integrated navigation system of the missile, the performance of data fusion algorithms based on the Cubature Kalman Filter (CKF) is seriously degraded when there are non-Gaussian noise and process-modeling errors in the system model. Therefore, a novel method is proposed, which is called Optimal Data Fusion algorithm based on the Adaptive Fading maximum Correntropy generalized high-degree CKF (AFCCKF-ODF). First, the Adaptive Fading maximum Correntropy generalized high-degree CKF (AFCCKF) is proposed and used as the local filter for the INS/GNSS and INS/CNS subsystems to improve the robustness of local state estimation. Then, the local state estimation is fused based on the minimum variance principle and high-degree cubature criterion to get the globally optimal state. Finally, the experimental results verify that the proposed algorithm can significantly improve the robustness of the missile-borne INS/CNS/GNSS integrated navigation system to non-Gaussian noise and process modeling error and obtain the global optimal navigation information.

Reducing Annotation Burden Through Multimodal Learning.

Choosing an optimal data fusion technique is essential when performing machine learning with multimodal data. In this study, we examined deep learning-based multimodal fusion techniques for the combined classification of radiological images and associated text reports. In our analysis, we (1) compared the classification performance of three prototypical multimodal fusion techniques: Early, Late, and Model fusion, (2) assessed the performance of multimodal compared to unimodal learning; and finally (3) investigated the amount of labeled data needed by multimodal vs. unimodal models to yield comparable classification performance. Our experiments demonstrate the potential of multimodal fusion methods to yield competitive results using less training data (labeled data) than their unimodal counterparts. This was more pronounced using the Early and less so using the Model and Late fusion approaches. With increasing amount of training data, unimodal models achieved comparable results to multimodal models. Overall, our results suggest the potential of multimodal learning to decrease the need for labeled training data resulting in a lower annotation burden for domain experts.

Multi-level Data Fusion Algorithm towards Privacy Protection in Wireless Sensor Networks

Data fusion is one of the key technologies in wireless sensor networks. To promote secure and efficient data fusion for wireless sensor networks, a privacy protection-based multi-level data fusion algorithm is proposed. In order to minimise the network energy consumption, the optimal number of cluster heads is selected and the adaptive clustering is performed according to the node energy and the positional relationship between nodes. Then, the cluster members collect and encrypt data, whereas the cluster head cleans and integrates the encrypted data. Furthermore, by analysing the correlation between data and constructing the BP neural network, cluster heads and the sink node can fuse the data in a cluster and the data between clusters to achieve the optimal data fusion. Results show that the mechanism proposed in this paper can significantly reduce resource overheads, effectively guarantee data security and fusion efficiency.

Track Occupation Detection Based on a Maximum Posterior Probability Model Using Multisensor Data Fusion

To address the railway track section occupation detection failure, this paper proposes a maximum posterior probability model that uses multisensor information fusion to detect track occupation. Based on the installation method of the sensor, this model obtains stable base data of occupied track sections and extracts its features, including train's running velocity, acceleration, direction, occupied area, wheelset axle counting, and track vibration. The maximum posterior probability and logarithm model are then derived by computing the prior probability, the posterior probability, and the conditional joint probability for the features. The judgment of the track occupation is more accurate compared with experience value. The experiments demonstrate that our track occupation detection method is able to effectively judge the occupation of the train, persons, and tool cart. Based on the maximum posterior probability, the Bayes optimal data fusion ratio for a measured parameter in this paper reaches 99.9983%.

Visible/near Infrared Reflection Spectrometer and Electronic Nose Data Fusion as an Accuracy Improvement Method for Portable Total Soluble Solid Content Detection of Orange

The visible/near infrared (VIS/NIR) spectrometer and electronic nose (E-nose) are two commonly used portable and nondestructive detection apparatuses which have a promising application for the quick acquisition of fruit’s internal quality in both the orchard and market. However, the accuracy of these instruments is sometimes unsatisfactory, especially for thick peeled fruit like the ‘Aiyuan 38’ orange, which was investigated in this research. The objective of this research was to find a method to improve the accuracy for the detection of an orange’s total soluble solid content (TSS) using a VIS/NIR spectrometer and E-nose. Different spectrum detection positions and conventional feature extraction methods are compared to get the optimal data fusion parameters. The detection model was then built up based on the obtained fusion data under the optimal parameters. Partial least squares regression (PLSR) and mutual information theory (MIT) were applied for feature extraction, and PLSR and principal component analysis (PCA)-back propagation neural network (BPNN) were applied for modeling and detection. PLSR results showed that the sampling reflection spectrum at the position of the calyx results in a better orange TSS detection than other sampling positions. For VIS/NIR reflection spectrum feature extraction, PLSR and MIT results showed that the optimal data process + feature extraction method is Savitzky-Golay + 763 features, when their mutual information values between the feature and TSS value were larger than 0.74. For E-nose feature extraction, PLSR and MIT results showed that the combined feature (combination of 75 s value, average value, average of differential value, integral value, and maximum value) is the optimal feature extraction method, and all features are retained for modeling. The PLSR detection ability of orange TSS based on fusion data is better than the single detection method, with the detection ability of the single detection methods being unsatisfactory. PCA-BPNN has better orange TSS detection ability than PLSR. The R2, RMSE, and slope from the calibration set for PCA-BPNN detection were 0.9695, 0.1895, and 0.9665, respectively, and from the validation set for PCA-BPNN detection were 0.8872, 0.4709, and 1.0871, respectively, indicating that this method can detect orange TSS efficiently.

Optimal fusion of industrial data streams with different granularities

In the Industry 4.0 era, it is a common situation that measurements are available with different resolutions (or granularities). Multiresolution (or multi-granularity) data can arise, for instance, from aggregation rules applied over measurements collected in batches during non-overlapping periods of time, or in composite sampling schemes (e.g., using industrial auto-sampler devices). Current solutions for optimal data fusion, such as Kalman filters (KF) and Moving Horizon Estimation (MHE) approaches, assume that all variables are measured at the same resolution, even if their acquisition rates are different. Therefore, they are unable to account for the multiresolution structure of data. In this article we introduce a multiresolution framework for the optimal fusion of multiple signals at distinct resolutions. This approach is compared against the benchmark alternative using simulated case studies, where the superior signal processing capability of the novel methodology is demonstrated.