Level Data Fusion Research Articles

Development of chemometric models using Vis-NIR and Raman spectral data fusion for assessment of infant formula storage temperature and time

This study evaluated the potential of Vis-NIR and Raman spectral data fusion combined with PLS and SVM chemometric models developed using a large dataset (n = 1700) of commercial infant formula (IF) samples to (i) discriminate between different IF storage temperature (20, 37 °C) and (ii) predict IF storage time (0–12 months). Three interval-based PLS variable selection methods (forward interval PLS (FiPLS), backward interval PLS (BiPLS) and synergy interval PLS (SiPLS)) and SVM-recursive feature elimination (SVM-RFE) methods were compared for model development. The best IF storage temperature discrimination model was developed using SVM classification (SVMC) and Vis-NIR spectra (400–2498 nm) (AccuracyCV = 99.82%, AccuracyP = 100%). SVM regression (SVMR) models developed using medium level data fusion (features selected by SVM-RFE) had the lowest root mean square error (RMSE) values for IF samples stored at either temperature, 20 °C or 37 °C (RMSECV = 0.7–0.8, RMSEP = 0.6–0.9). Industrial relevanceSpectroscopic technologies, including Vis-NIR and Raman spectroscopy have been widely applied for process analysis and increasingly for on-line process monitoring in areas of chemicals, food processing, agriculture and pharmaceuticals, etc. Due to their rapid measurement and minimal or no sample preparation, they are highly suitable for in-line process monitoring. This study demonstrates that Vis-NIR and Raman process analytical tools either individually or combined may be employed for quality assessment and process control of IF manufacture.

Complement component face space for 3D face recognition from range images

This paper proposes a mathematical model for decomposing a range face image into four basic components (named ‘complement components’) in conjunction with a simple approach for data-level fusion to generate thirty-six additional hybrid components. These forty component faces composing a new face image space called the ‘complement component face space.’ The main challenge of this work was to extract relevant features from the vast face space. Features are extracted from the four basic components and four selected hybrid components using singular value decomposition. To introduce diversity, the extracted feature vectors are fused by applying the crossover operation of the genetic algorithm using a Hamming distance-based fitness measure. Particle swarm optimization-based feature selection is employed on the fused features to discard redundant feature values and to maximize the face recognition performance. The recognition performances of the proposed feature set with a support vector machine-based classifier on three accessible and well-known 3D face databases, namely, Frav3D, Bosphorus, and Texas3D, show significant improvements over those achieved by state-of-the-art methods. This work also studies the feasibility of utilizing the component images in the complement component face space for data augmentation in convolutional neural network (CNN)-based frameworks.

Data fusion-based sustainable digital twin system of intelligent detection robotics

In order to give full play to the role of science and technology in achieving the sustainable development goals, digital twin technology based on data fusion plays an irreplaceable role. This paper analyzes the detection process of intelligent detection robots for massage chairs, theoretical research is carried out from two aspects of decision-level fusion and data-level fusion. The principles, levels, and models of information fusion are analyzed, and an in-depth analysis and discussion of DS evidence theory are conducted.Whereas the ambiguity and uncertainty of evaluation indicators in sustainable design, a quantitative model for sustainable design is established in this paper. The entropy weight method is used to determine the weight of sub-indicators, and a decision matrix for the indicators to be evaluated is established. Based on the data detected by the robot while the massage chair is working, the sustainability of the massage chair is evaluated. Reflect the obtained quantitative indicators on the digital twin data flow system, improve the comprehensive performance of products in sustainable design through data fusion, and prove that the necessity of sustainability and the proposed method must be considered in the design and use process feasibility.

SPECTRAL ALGORITHM FOR CONTENT-BASED IMAGE RETRIEVAL

Colour images are rich in visual information. The process of searching for the most similar images in large-scale database based on visual features of query image is still a challenge in Content-Based Image Retrieval (CBIR) due to a semantic gap issue. In this paper, we proposed a fusing retrieval method to diminish the gap between high-level and low-level meanings by involving two aspects. The first aspect is increasing the effectiveness of image representation. Hence, data-level fusion features were suggested, a local feature from Discrete Cosine Transform (DCT) and Local Binary Patterns (LBP) in frequency and spatial domains respectively that was applied by a spectral clustering algorithm (graph-based) in addition to a global weighted LBP feature. The second aspect is fusing multiple retrieved similarity measures (scores/evidences) obtained from above global (LBP) and local features (DCTLBP) in terms of score-level fusion. The method is evaluated in WANG standard publically dataset.

RUL Prediction and Uncertainty Management for Multisensor System Using an Integrated Data-Level Fusion and UPF Approach

Due to the fact that single sensor data contains only partial information about complex systems, multiple sensors are often embedded to simultaneously monitor the health state and predict the remaining useful life (RUL). This brings new challenges to traditional approaches that focus on single sensor in terms of data fusion, optimization, and uncertainty management. To address these challenges, this article proposes a novel RUL prediction and uncertainty management framework for multisensor systems. In this framework, a composite 1-D health indicator (1-D HI) is obtained from multiple sensors by optimizing some HI characteristics, including monotonicity, robustness, fitting error, and range information, to better describe the underlying degradation process. A multiobjective grasshopper optimization algorithm is used to achieve the optimal weight vector of the fusion model. Then, an unscented particle filter is introduced to predict the RUL by combining the degradation model constructed from HI and the composite 1-D HI as measurement. To manage the uncertainty in prognosis, a probability distribution of failure threshold and noise parameter adjustment are developed. Experimental results on aircraft turbine engine degradation and comparison with state-of-the-art methods are presented to demonstrate the effectiveness of the proposed framework in RUL prediction of multisensor systems.

Sensor data fusion and clustering : A congestion detection and avoidance approach in wireless sensor networks

A wireless sensor network is gaining more popularity as it is easy to maintain, manage and inexpensive to establish. A group of sensor nodes with sensing and computing capability are deployed in network. When these nodes are deployed in remote areas, an adequate energy of the nodes becomes major issue. As the sensor nodes are battery operated, so energy is an important resource that must be utilized efficiently. Therefore, the most preferable method to tackle such issue efficiently is to divide the network into clusters and data communication by data fusion techniques. Data fusion is a technique which enables gathering data from several sensor nodes thus to provide a unified scenario in order to reduce redundancy of data which consumes lots of energy. The proposed method is clustered based data fusion to avoid congestion. One node from each cluster is elected as cluster head which is elected after observing various parameters like enduring energy of the node, distance from base station, and chances of traffic in near future. The cluster head node gathers data from cluster members hence applies local level data fusion on this information and then forward it to base station for final decision level data fusion. This method improves the reliability of data, reduced traffic and improve network link. The level of congestion is predicted periodically. It is taken as congestion indicator. This method helps in reducing power consumption and packet drop rate. The performance of this technique is investigated to prove its superiority to different state of art algorithms. The performance is validated using numerical analysis derivations and MATLAB.

Two Level Data Fusion Model for Data Minimization and Event Detection in Periodic Wireless Sensor Network

Two Level Data Fusion Model for Data Minimization and Event Detection in Periodic Wireless Sensor Network

Data fusion strategies for energy efficiency in buildings: Overview, challenges and novel orientations

Recently, tremendous interest has been devoted to develop data fusion strategies for energy efficiency in buildings, where various kinds of information can be processed. However, applying the appropriate data fusion strategy to design an efficient energy efficiency system is not straightforward; it requires a priori knowledge of existing fusion strategies, their applications and their properties. To this regard, seeking to provide the energy research community with a better understanding of data fusion strategies in building energy saving systems, their principles, advantages, and potential applications, this paper proposes an extensive survey of existing data fusion mechanisms deployed to reduce excessive consumption and promote sustainability. We investigate their conceptualizations, advantages, challenges and drawbacks, as well as performing a taxonomy of existing data fusion strategies and other contributing factors. Following, a comprehensive comparison of the state-of-the-art data fusion based energy efficiency frameworks is conducted using various parameters, including data fusion level, data fusion techniques, behavioral change influencer, behavioral change incentive, recorded data, platform architecture, IoT technology and application scenario. Moreover, a novel method for electrical appliance identification is proposed based on the fusion of 2D local texture descriptors, where 1D power signals are transformed into 2D space and treated as images. The empirical evaluation, conducted on three real datasets, shows promising performance, in which up to 99.68% accuracy and 99.52% F1 score have been attained. In addition, various open research challenges and future orientations to improve data fusion based energy efficiency ecosystems are explored.

Multi-Sensor Data Fusion Algorithm Based on BP Neural Network

In multi-sensor detection system, the application of multi-sensor accurate detection system parameters is limited due to the existence of measurement noise. Using multi-source data fusion technology can be more accurate, timely detection and data processing system. Data fusion is a basic function in humans and other biological systems. In this paper, in order to make the system adaptive multi-source data fusion, using the BP neural network algorithm is a good way to deal with incomplete test data and test the noise problem. In this paper, the characteristics of three levels of data fusion and the derivation process of BP neural network algorithm are introduced in detail. In order to verify the role of BP neural network algorithm in the process of detection system filtering, a MATLAB simulation experiment is carried out. The experimental results show that the BP neural network algorithm can effectively reduce the measurement error of multi-sensor detection system and improve the detection accuracy.

A new analytical method for discrimination of species in Ganodermataceae mushrooms

ABSTRACT A new analytical approach for the species discrimination of Ganodermataceae mushroom was developed by using data fusion strategy based on attenuated total reflectance Fourier transform infrared ATR-FTIR and ultraviolet–visible (UV–vis) spectroscopy, and applying the chemometric tools. The optimization for determination of UV–vis spectra was described. The multivariate discrimination ways used were t-distributed Stochastic Neighbor Embedding (t-SNE), Partial Least Squares-Discriminant Analysis (PLS-DA), and Random Forest (RF). The data fusion levels used were low- and mid-data fusion. The performance of the model was assessed by several parameters as root mean square error of estimation (RMSEE), root mean square error of cross validation (RMSECV), R2Y(cum), and Q2 (cum). The discrimination results were evaluated by accuracy from the test set, which was composed of samples with unknown origin. The new proposed method took shorter time and lower cost, and the results showed good discrimination power among various species. PLS-DA and RF models based on mid-level fusion data were able to classify mushrooms according to real origin, confirming the potential of data fusion and chemometrics in mushrooms species discrimination.

A Comparative Study of Five Networks for Reservoir Classification Based on Geophysical Logging Signals

Unconventional reservoir classification suffers low accuracy because of the complex geophysical properties. With good performance and moderate cost, geophysical logging is considered to be of great potential as the compromise solution between seismic and core. Since the “sweet-spot” depending on the sands' properties are implied not in the logging “point” but “segment”, 1D-CNN is supposed to be a better fit for reservoir classification: It doesn't pay too much attention to the logging signal sequence itself like LSTM, but focuses on feature extraction from the whole signals combination. Moreover, samples would have various sizes because of different layer thickness. While 1D-CNN requires all samples must be converted to a uniform size before input. 1D fully convolutional network (1D-FCN) can receive any size input of layer thickness without manually aligning it to the same size. For the first time in this paper, structures of the five networks including two fully connected networks (global mapping artificial neural network, GM-ANN and point-to-point mapping artificial neural network, PPM-ANN), two fully convolutional networks (1D fully convolutional network with decision-level fusion, 1D-FCN-DEF and 1D fully convolutional network with data-level fusion, 1D-FCN-DAF) and the common 1D convolutional neural network (1D-CNN) are compared and evaluated the suitability for processing logging data in detail. Results on tight gas in Ordos basin of China illustrated by receiver operating characteristics (ROC) curves show that 1D-FCN-DEF and 1D-FCN-DAF have achieved the higher area under the curve (AUC) with the values of 0.8889 and 0.9107 respectively in average comparing to 0.7515, 0.8006 and 0.8364 of GM-ANN, PPM-ANN and 1D-CNN. Case study proves that 1D-FCNs are more accurate than other networks. This paper provides a suitable new idea for logging interpretation and expands the application scope of DL in geophysical logging signals processing.

Optimal estimation of low-rank factors via feature level data fusion of multiplex signal systems

The design of fusion engines is a subject of great importance in a variety of fields. In this paper, we focus on the problem of linear fusion at the feature level for multiple signal matrices with noises, with the features being extremal eigenvectors. When given multiple similarity matrices, the objective is to find an estimate of the latent signal eigenspace. The concentration result for the inner product of features from different matrix samples is developed, utilizing the random matrix theory. Based on of the theoretical results, we proposed an efficient algorithm, EigFuse, to solve the constrained data-driven optimization problem with different level of noises. Our method is of high efficiency by comparing it with state-of-the-art baseline approaches with multiple noise levels. Comprehensive experiments on several synthetic as well as real-life networks demonstrate our method's superior performance.

Exploration of Chinese Sign Language Recognition Using Wearable Sensors Based on Deep Belief Net

In this paper, deep belief net (DBN) was applied into the field of wearable-sensor based Chinese sign language (CSL) recognition. Eight subjects were involved in the study, and all of the subjects finished a five-day experiment performing CSL on a target word set consisting of 150 CSL subwords. During the experiment, surface electromyography (sEMG), accelerometer (ACC), and gyroscope (GYRO) signals were collected from the participants. In order to obtain the optimal structure of the network, three different sensor fusion strategies, including data-level fusion, feature-level fusion, and decision-level fusion, were explored. In addition, for the feature-level fusion strategy, two different feature sources, which are hand-crafted features and network generated features, and two different network structures, which are fully-connected net and DBN, were also compared. The result showed that feature level fusion could achieve the best recognition accuracy among the three fusion strategies, and feature-level fusion with network generated features and DBN could achieve the best recognition accuracy. The best recognition accuracy realized in this study was 95.1% for the user-dependent test and 88.2% for the user-independent test. The significance of the study is that it applied the deep learning method into the field of wearable sensors-based CSL recognition, and according to our knowledge it's the first study comparing human engineered features with the network generated features in the correspondent field. The results from the study shed lights on the method of using network-generated features during sensor fusion and CSL recognition.

Using Hellinger and Bures metrics to construct two-dimensional quantum metric space for weather data fusion

Motivated by obtaining more concise and reliable fusion results for relevant applications, such as, in sector monitoring, risk managing, early-warning, dynamic prediction, this paper explores the detection and fusion models of weather data from multiple perspectives. We represent the weather data as the collection of quantum states, and study construction of the quantum metric space in which the related weather data fusion method is developed. Since weather data usually contain raw (image or sensor) data, feature data, or decision data, the weather data fusion belongs to normal three levels of data fusion, i.e., pixel level, feature level or decision level. In this paper we represent weather data units as density matrixes, and take the quantified data units as the basis for constructing the two-dimensional quantum metric space consisting of Hellinger and Bures metrics. All the density matrix units in this space are transformed into the rectangle nodes for detection and fusion. Then, according to the neighborhood relationships between rectangle nodes and the density of rectangle nodes in a specified rectangle, the source weather dataset is divided into different subsets. The key to the fusion of the rectangle nodes in a subset is to calculate the presupposing central node in this subset's corresponding rectangle. And the corresponding object node depends on the areas of the rectangles including rectangle nodes and the central node. The experimental evaluation demonstrates that, compared with the already developed fusion methods, the proposed weather data fusion method can obtain more concise and reliable fusion results for decision applications.

Performance Improvement of Classifier in Fault Diagnosis of Rotating Machines Using Sensor Fusion Techniques

The shaft, rotor, bearing and gear are the important elements of the rotating machines. Most of the problems in rotating machines are caused due to bearings and shaft. The failure of rotating machine causes production downtime and economic & safety issues. Vibration signal analysis is highly accepted technique in fault diagnosis of rotating machine. For automation of fault diagnosis, machine learning approach has been followed. Machine learning classifies fault based on variation in signatures pattern of the machine. But its effectiveness gets reduced when it is used for multi fault class problem. So in the present work, sound signals are also used along with vibration signals for applying sensor fusion techniques. In sensor fusion, signals from various sensors are fused in three levels such as data fusion, feature fusion and decision level fusion and the fused data sets are used for fault classification using machine learning algorithm. The performance of each technique is studied in detail and compared using classification accuracy. A new method is proposed by combination of fusion techniques to enhance the performance.

A holistic approach to authenticate organic sweet oranges (Citrus Sinensis L. cv Osbeck) using different techniques and data fusion

Spain is the second largest European producer of oranges and the fourth in the world. Moreover, in Andalusia, the land dedicated to organic farming of oranges has doubled in the last decade (2006–2016). The great challenge in the organic farming sector is to find a reliable analytical technique (or a set thereof) to verify that the method of production followed the organic production guidelines. A new methodology based on stable isotopes, (poly)phenol profiling and conventional quality data in the edible and non-edible parts of the oranges was developed using data fusion techniques. This system allowed authenticating organic fresh oranges. The application of a low level data fusion and a non-linear algorithm (k-nearest neighbors), obtained 100% specificity and sensitivity in the test set models. This work presented a new focus on dealing with the problem of the authentication of fruits grown under organic farming techniques.

Traffic speed prediction for intelligent transportation system based on a deep feature fusion model

Currently, many types of traffic data from different advanced data collection techniques are available. Plenty of effort has been spent to take full advantage of the heterogeneous data to enhance the prediction accuracy of the model in the advanced travel information system. The objective of this study is to build a deep feature fusion model to predict space–mean–speed using heterogeneous data. The temporal and spatial features are defined as the raw input which are extracted and trained by stacked autoencoders in the first step. Then, the extracted representative features from the data are fused. Finally, prediction models can be developed to capture the correlations. Therefore, the prediction model can consider both temporal–spatial correlation and correlation of heterogeneous data. Using the real-world data, some machine learning models including artificial neural network, support vector regression, regression tree and k-nearest neighbor are implemented and compared. The best result can be obtained when deep feature fusion model and support vector regression are jointly applied. Moreover, we compare the new proposed deep feature–level fusion method with the widely used data–level fusion method. The results indicate that proposed deep feature fusion model can achieve a better performance.

Hierarchical data fusion for Smart Healthcare

The Internet of Things (IoT) facilitates creation of smart spaces by converting existing environments into sensor-rich data-centric cyber-physical systems with an increasing degree of automation, giving rise to Industry 4.0. When adopted in commercial/industrial contexts, this trend is revolutionising many aspects of our everyday life, including the way people access and receive healthcare services. As we move towards Healthcare Industry 4.0, the underlying IoT systems of Smart Healthcare spaces are growing in size and complexity, making it important to ensure that extreme amounts of collected data are properly processed to provide valuable insights and decisions according to requirements in place. This paper focuses on the Smart Healthcare domain and addresses the issue of data fusion in the context of IoT networks, consisting of edge devices, network and communications units, and Cloud platforms. We propose a distributed hierarchical data fusion architecture, in which different data sources are combined at each level of the IoT taxonomy to produce timely and accurate results. This way, mission-critical decisions, as demonstrated by the presented Smart Healthcare scenario, are taken with minimum time delay, as soon as necessary information is generated and collected. The proposed approach was implemented using the Complex Event Processing technology, which natively supports the hierarchical processing model and specifically focuses on handling streaming data ‘on the fly’—a key requirement for storage-limited IoT devices and time-critical application domains. Initial experiments demonstrate that the proposed approach enables fine-grained decision taking at different data fusion levels and, as a result, improves the overall performance and reaction time of public healthcare services, thus promoting the adoption of the IoT technologies in Healthcare Industry 4.0.

Fusing spectral and textural information in near-infrared hyperspectral imaging to improve green tea classification modelling

Hyperspectral imaging (HSI) can acquire data in two modes: imaging and spectroscopy, revealing the spatially-resolved spectral properties of materials. Traditional HSI processing in the close-range domain primarily focuses on the spectral information with minimal utilisation of the spatial information present in the data. The present work describes a methodology for utilising the spatial information present in HSI data to improve classification modelling over that achievable with spectral information alone. The methodology has been evaluated using near infrared (NIR) HSI data of sixteen green tea products from seven different countries. The methodology involves selecting and sharpening an image plane to enhance the textural details. The textural information is then extracted from the statistical properties of the grey level co-occurrence matrix (GLCM) of the sharpened image plane using a moving window operation. Finally, the textural properties are combined with the spectral information using one of the three different levels of data fusion, i.e. raw data level, feature level and decision level. Raw data-level fusion involved concatenating the spectral and textural data before performing the classification task. The feature-level fusion involved performing principal component analysis (PCA) on spectral and textural information and combining the PC scores obtained prior to performing classification. Decision-level fusion involved a majority voting scheme to enhance the final classification maps. All the classification tasks were performed using multi-class support vector machine (SVM) models. The results showed that combining the textural and spectral information during modelling resulted in improved classification of the sixteen green tea products compared to models built using spectral or textural information alone.

Multiple Sensors Based Prognostics With Prediction Interval Optimization via Echo State Gaussian Process

In prognostics and health management, multiple sensors have been widely used to monitor the health condition of complex machines. As each sensor provides partial and dependent information, data-level fusion techniques have been developed and aim at increasing the reliability and safety of machines by providing the current health assessment and remaining useful life (RUL) prediction. While most existing data-level fusion techniques have shown a promise for prognostics, they are mainly limited by only focusing on the improvement of point prediction, which makes it difficult to provide adequate available information for decision-making in predictive maintenance. In this paper, the prediction interval (PI) is adopted to model the prediction uncertainty of RUL for its nature of high variability. An improved echo state Gaussian process (IESGP), which is able to provide an estimation of prediction uncertainty, is developed as a novel data-driven approach for RUL prediction and PI construction. The proposed IESGP is a novel Bayesian approach which combines the merits of echo state networks and Gaussian processes to enhance the prediction accuracy. Based on this, a comprehensive cost function is constructed to characterize the accuracy and uncertainty simultaneously, then a multi-objective genetic algorithm (MOGA) is applied to optimize the point prediction and PI. The validity of the proposed approach is verified on the widely used turbofan benchmark datasets. The experimental results show that the proposed approach can not only achieve superior predictive accuracy in comparison with several state-of-the-art approaches but also obtain PIs with high quality.