Search

Radar and camera information fusion sensing methods are used to solve the inherent shortcomings of the single sensor in severe weather. Our fusion scheme uses radar as the main hardware and camera as the auxiliary hardware framework. At the same time, the Mahalanobis distance is used to match the observed values of the target sequence. Data fusion based on the joint probability function method. Moreover, the algorithm was tested using actual sensor data collected from a vehicle, performing real-time environment perception. The test results show that radar and camera fusion algorithms perform better than single sensor environmental perception in severe weather, which can effectively reduce the missed detection rate of autonomous vehicle environment perception in severe weather. The fusion algorithm improves the robustness of the environment perception system and provides accurate environment perception information for the decision-making system and control system of autonomous vehicles.

With increasing usage of face recognition algorithms, it is well established that external artifacts and makeup accessories can be applied to different facial features such as eyes, nose, mouth, and cheek, to obfuscate one’s identity or to impersonate someone else’s identity. Recognizing faces in the presence of these artifacts comprises the problem of disguised face recognition, which is one of the most arduous covariates of face recognition. The challenge becomes exacerbated when disguised faces are captured in real-time environment, with low resolution images. To address the challenge of disguised face recognition, this paper first proposes a novel multi-objective encoder-decoder network, termed as DED-Net. DED-Net attempts to learn the class variations in the feature space generated by both disguised as well non-disguised images, using a combination of Mahalanobis and Cosine distance metrics, along with Mutual Information based supervision. The DED-Net is then extended to learn from the local and global features of both disguised and non-disguised face images for efficient face recognition, and the complete framework is termed as Disguise Resilient (D-Res) framework. The efficacy of the proposed framework has been demonstrated on two real-world benchmark datasets: Disguised Faces in the Wild (DFW) 2018 and DFW2019 competition datasets. In addition, this research also emphasizes on the importance of recognizing disguised faces in low resolution settings and proposes three experimental protocols to simulate the real-world surveillance scenario. To this effect, benchmark results have been shown on seven protocols for three low resolution settings (32×32, 24×24, and 16×16) of the two DFW benchmark datasets. The results demonstrate superior performance of the D-Res framework, in comparison with benchmark algorithms. For example, an improvement of around 3% is observed on the Overall protocol of the DFW2019 dataset, where the D-Res framework achieves 96.3%. Experiments have also been performed on benchmark face verification datasets (LFW, YTF, and IJB-B), where the D-Res framework achieves improved verification accuracy.

Inertial Navigation System/Celestial Navigation System (INS/CNS) integration, especially for the tightly-coupled mode, provides a promising autonomous tactics for Hypersonic Vehicle (HV) in military demands. However, INS/CNS integration is a challenging research task due to its special characteristics such as strong nonlinearity, non-additive noise and dynamic complexity. This paper presents a novel nonlinear filtering method for INS/CNS integration by adopting the emerging Cubature Kalman Filter (CKF) to handle the strong INS error model nonlinearity caused by HV’s high dynamics. It combines the state-augmentation technique into the nonlinear CKF to decrease the negative effect of non-additive noise in inertial measurements. Subsequently, a technique for the detection of dynamic model uncertainty is developed, and the augmented CKF is modified with fading memory to tackle dynamic model uncertainty by rigorously deriving the fading factor via the theory of Mahalanobis distance without artificial empiricism. Simulation results and comparison analysis prove that the proposed method can effectively curb the adverse impacts of non-additive noise and dynamic model uncertainty for inertial measurements, leading to improved performance for HV navigation with tightly-coupled INS/CNS integration.

To simultaneously monitor some electrical or mechanical faults of an in-wheel motor and intelligently evaluate the operation safety, a fuzzy system of operation safety assessment (OSA) is proposed. This method firstly uses many symptom parameters (SPs) such as root mean square, crest factor, temperature rise and current covariance to express the features of the electrical and mechanical faults from different perspectives such as vibration, noise, temperature, current and voltage, possibility theory is employed to translate the probability density function of each SP into the possibility function, and sample data are gradually updated to optimize the possibility function for obtaining the SPs’ membership functions that are evaluation models. Secondly, the probabilities of the current operation state that is safety, attention or danger are obtained from each evaluation model in a stage. Picture fuzzy set (PFS) is used to define a basic picture fuzzy number (PFN), then many PFNs from multiple models and multiple stages are used to establish an OSA's decision matrix. Thirdly, Mahalanobis distance is reintegrated into PFS's theory for objectively judging the real-time evaluation information, and best-worst method is used to estimate subjectively the initial evaluation experience, then the multi-model linkage mechanism is designed. Finally, TODIM is modified to define the relative safety ratio, and prospect theory is employed to structure the global index for formulating the multi-stage collaboration approach, then a fuzzy OSA's system is established. The effectiveness of the proposed method was verified by experimental analysis for the operation safety of in-wheel motor with electrical and mechanical faults.

Face verification can be regarded as a two-class fine-grained visual-recognition problem. Enhancing the feature's discriminative power is one of the key problems to improve its performance. Metric-learning technology is often applied to address this need while achieving a good tradeoff between underfitting, and overfitting plays a vital role in metric learning. Hence, we propose a novel ensemble cascade metric-learning (ECML) mechanism. In particular, hierarchical metric learning is executed in a cascade way to alleviate underfitting. Meanwhile, at each learning level, the features are split into nonoverlapping groups. Then, metric learning is executed among the feature groups in the ensemble manner to resist overfitting. Considering the feature distribution characteristics of faces, a robust Mahalanobis metric-learning method (RMML) with a closed-form solution is additionally proposed. It can avoid the computation failure issue on an inverse matrix faced by some well-known metric-learning approaches (e.g., KISSME). Embedding RMML into the proposed ECML mechanism, our metric-learning paradigm (EC-RMML) can run in the one-pass learning manner. The experimental results demonstrate that EC-RMML is superior to state-of-the-art metric-learning methods for face verification. The proposed ECML mechanism is also applicable to other metric-learning approaches.

We propose a novel constraint-handling technique for the covariance matrix adaptation evolution strategy (CMA-ES). The proposed technique is aimed at solving explicitly constrained black-box continuous optimization problems, in which the explicit constraint is a constraint whereby the computational time for the constraint violation and its (numerical) gradient are negligible compared to that for the objective function. This method is designed to realize two invariance properties: invariance to the affine transformation of the search space, and invariance to the increasing transformation of the objective and constraint functions. The CMA-ES is designed to possess these properties for handling difficulties that appear in black-box optimization problems, such as non-separability, ill-conditioning, ruggedness, and the different orders of magnitude in the objective. The proposed constraint-handling technique (CHT), known as ARCH, modifies the underlying CMA-ES only in terms of the ranking of the candidate solutions. It employs a repair operator and an adaptive ranking aggregation strategy to compute the ranking. We developed test problems to evaluate the effects of the invariance properties, and performed experiments to empirically verify the invariance of the algorithm. We compared the proposed method with other CHTs on the CEC 2006 constrained optimization benchmark suite to demonstrate its efficacy. Empirical studies reveal that ARCH is able to exploit the explicitness of the constraint functions effectively, sometimes even more efficiently than an existing box-constraint handling technique on box-constrained problems, while exhibiting the invariance properties. Moreover, ARCH overwhelmingly outperforms CHTs by not exploiting the explicit constraints in terms of the number of objective function calls., Accepted to Evolutionary Computation (MIT Press)

Wind turbines are complex systems that are susceptible to frequent anomalies, faults, and abnormal behaviour. These are caused mainly due to off-nominal conditions, catastrophic events, and major failures, resulting in downtime conditions. An accurate and timely detection of downtime events provides crucial information for planning and decision-making. This study investigates the utility of wind power and wind speed as potential parameters for real-time downtime detection. Early and accurate detection of these anomalies using system outputs collected from monitoring stations is challenging and involved, especially when attempted in real-time. In this article, a real-time downtime detection framework is proposed that maps system outputs to turbine events - faults, scheduled, and unplanned maintenance - through online condition indicators. Without imposing strong distributional assumptions, using available training samples, an optimal, cost-sensitive real-time anomaly detection framework is proposed to determine whether a sample is anomalous. Considering the trade-off between misclassification errors and detection rates, detection studies are performed using wind power and speed - calibrated against available alarm classifiers - obtained from two Irish wind farms. The data cleaning and formatting for analysis was automated and subjected to classification with several levels of complexity. Recursive condition indicators (RCIs) such as Recursive Mahalanobis distance (RMD) and Recursive Residual Error (RRE) are chosen as features for classification. The real-time detection model becomes particularly useful when it is prohibitive to identify in advance the anomalies without a baseline of the system behaviour under such conditions. Case studies involving Irish wind Supervisory Control And Data Acquisition (SCADA) data demonstrate the successful application of the proposed work for early and accurate downtime detection with comparison to a reference machine learning approach.

This paper highlights kernel principal component analysis (KPCA) in distinguishing damage-sensitive features from the effects of liquid loading on frequency response. A vibration test is performed on an aircraft wing box incorporated with a liquid tank that undergoes various tank loading. Such experiment is established as a preliminary study of an aircraft wing that undergoes operational load change in a fuel tank. The operational loading effects in a mechanical system can lead to a false alarm as loading and damage effects produce a similar reduction in the vibration response. This study proposes a non-nonlinear transformation to separate loading effects from damage-sensitive features. Based on a baseline data set built from a healthy structure that undergoes systematic tank loading, the Gaussian parameter is measured based on the distance of the baseline data set to various damage states. As a result, both loading and damage features expand and are distinguished better. For novelty damage detection, Mahalanobis square distance (MSD) and Monte Carlo-based threshold are applied. The main contribution of this project is the nonlinear PCA projection to understand the dynamic behavior of the wing box under damage and loading influences and to differentiate both effects that arise from the tank loading and damage severities.

As the number of cloud services (CSs) offering similar functionality is growing, more attention has been payed on the quality of service (QoS) of CSs. However, in a dynamic cloud environment, the explicit and inherent variation of QoS causes the single CS selection via collaborative filtering techniques (CSS-CFT) to be challengeable. A variation-aware approach via collaborative QoS prediction is proposed to select an optimal CS according to users' non-functional requirements. Based on time series QoS data, this approach utilizes a set of specific cloud models to quantify the variation characteristics of QoS from the four aspects including central tendency, variation range, frequency of variation and period. To exactly identify the neighboring users for a current user, this paper employs the double Mahalanobis distances to measure the similarity of QoS cloud models. The variation-aware CSS-CFT is formulated as a multi-criteria decision-making problem, and an improved TOPSIS method is exploited to solve it, by considering both the objective QoS variation and subjective user preferences during different time periods. The experiments based on a real-world dataset demonstrate that the proposed approach can enhance the accuracy of CSS-CFT in a high-variance environment without noticeable increase of selection time, in comparison to the existing approaches.

The microstructure is the key factor for quality discriminate of coke. In view of the characteristics of coke optical tissue (COT), a segmentation method of coke microstructures based on adaptive clustering was proposed. According to the strategy of multiresolution, adaptive threshold binarization and morphological filtering were carried out on COT images with lower resolution. The contour of the COT body was detected through the relationship checking between contours in the binary image, and hence, COT pixels were picked out to cluster for tissue segmentation. In order to get the optimum segmentation for each tissue, an advanced K -means method with adaptive clustering centers was provided according to the Calinski-Harabasz score. Meanwhile, Euclidean distance was substituted with Mahalanobis distance between each pixel in HSV space to improve the accuracy. The experimental results show that compared with the traditional K -means algorithm, FCM algorithm, and Meanshift algorithm, the adaptive clustering algorithm proposed in this paper is more accurate in the segmentation of various tissue components in COT images, and the accuracy of tissue segmentation reaches 94.3500%.

In public scenes such as stations and hospitals, the crowds are intensive and abnormal pedestrian often causes group hazards. The recognition of abnormal pedestrian is an important security problem, which is generally solved by inspection robots. Traditional visual feature methods pay much attention to the inherent attributes of pedestrians (such as gender and age), which ignores the complex semantic information displayed by pedestrian trajectories. This article uses scene monitoring visual sensors to analyze pedestrian trajectories in public scenes. We propose an abnormal trajectory recognition framework, which analyzes the pedestrian trajectories from clusters, deviation and trajectory entropy. In this framework, the convergence condition of the K-Means method is optimized to cluster the pedestrian destinations and trajectories; the Mahalanobis distance is used to evaluate the trajectory deviation; the dimensional feature is established through the velocity and angle difference of the trajectory. In the end, the results can prove that the methods in this article can successfully identify abnormal pedestrians.

Many industrial processes are equipped with a large number of sensors, which usually generate multichannel high-dimensional profiles that can be used to monitor the health condition and detect anomalies of the processes. However, the data irregularity, information obscurity, complex correlations, and nonlinear structures of the multichannel data pose significant challenges for the development of anomaly detection methodologies. To address these challenges, this article proposes a method, Mahalanobis Distance-based Functional Derivative Support Vector Data Description (MD-FDSVDD), for the process monitoring of applications with multichannel profiles. The proposed method first estimates a smooth function of each profile from its irregularly acquired observations and then takes its derivative function to enhance the characteristics associated with anomalies. Next, the smoothed derivative functions are transformed based on Mahalanobis distance to address the strong linear correlation challenge. Finally, the transformed derivative data are used to construct a functional SVDD model to detect anomalies. The effectiveness of the proposed method is evaluated using a simulated dataset and a real-world dataset from a Basic Oxygen Furnace steelmaking process.

Fault prediction in the analog circuits is a serious problem to be addressed on an immediate basis, as traditionally, the faults in the analog circuits are diagnosed only after their occurrence. Since the outcome of the faults creates highly expensive scenarios in case of the analog circuit industry, there is a need for an effective prediction model that keeps track of the faults prior to their occurrence. Accordingly, this article focuses on the fault prediction model in analog circuits using proposed deep model called, Rider-deep-long short-term memory (LSTM). Here, the significance and precision of the prediction relies on the fault indicator, which is computed based on three distance measures, such as mahalanobis distance, Euclidean distance, and angular distance, and thereby, enables an effective health estimation of the circuit. The estimation is effectively solved using the Rider-deep-LSTM, which is the integration of proposed Rider-Adam algorithm in deep-LSTM, for training the model parameters. The proposed prediction method acquires the Pearson correlation coefficient of 0.9973 and 0.9919 while using the circuits, such as solar power converter and low noise bipolar transistor amplifier.

We propose a new approach to the construction of $$k $$ -means clustering algorithms in which the Mahalanobis distance is used instead of the Euclidean distance. The approach is based on minimizing differentiable estimates of the mean insensitive to outliers. Illustrative examples convincingly show that the proposed algorithm is highly likely to be robust with respect to a large amount of outliers in the data.

Fetoscopic laser ablation surgery could be guided using a high-quality panorama of the operating site, representing a map of the placental vasculature. This can be achieved during the initial inspection phase of the procedure using image mosaicking techniques. Due to the lack of camera calibration in the operating room, it has been mostly modelled as an affine registration problem. While previous work mostly focuses on image feature extraction for visual odometry, the challenges related to large-scale reconstruction (re-localisation, loop closure, drift correction) remain largely unaddressed in this context. This letter proposes using pose graph optimisation to produce globally optimal image mosaics of placental vessels. Our approach follows the SLAM framework with a front-end for visual odometry and a back-end for long-term refinement. Our front-end uses a recent state-of-the-art odometry approach based on vessel segmentation, which is then managed by a key-frame structure and the bag-of-words (BoW) scheme to retrieve loop closures. The back-end, which is our key contribution, models odometry and loop closure constraints as a pose graph with affine warpings between states. This problem in the special Euclidean space cannot be solved by existing pose graph algorithms and available libraries such as G2O. We model states on affine Lie group with local linearisation in its Lie algebra. The cost function is established using Mahalanobis distance with the vectorisation of the Lie algebra. Finally, an iterative optimisation algorithm is adopted to minimise the cost function. The proposed pose graph optimisation is first validated on simulation data with a synthetic trajectory that has different levels of noise and different numbers of loop closures. Then the whole system is validated using real fetoscopic data that has three sequences with different numbers of frames and loop closures. Experimental results validate the advantage of the proposed method compared with baselines.

Data samples of complicated geometry and nonlinear separability are considered as common challenges to clustering algorithms. In this article, we first construct Mahalanobis distance in the kernel space and then propose a novel fuzzy clustering model with a kernelized Mahalanobis distance, namely KMD-FC. The key contributions of KMD-FC include: first, the construction of KMD matrix is innovatively transformed from the Euclidean distance kernel matrix, which is able to effectively avoid the problem of “curse of dimensionality” posed by explicitly calculating the sample covariance matrix in the kernel space; second, for the first time, the kernelized Gustafson–Kessel (GK) fuzzy C-means algorithm is achieved, which is critically important to extend the applications of the GK algorithm to the nonlinear classification tasks; finally, taking account of the overall distribution of samples in the kernel space after kernel mapping to improve the generalizability of the proposed KMD-FC clustering method. Comprehensive experiments conducted on a wide range of datasets, including synthetic datasets and machine learning repository (UCI) datasets, have validated that the proposed clustering algorithm outperformed the state-of-the-art methods in comparison.

Anomaly Detection with Hyper Spectral Image (HSI) refers to finding a significant difference between the background and the anomalous pixels present in the image. This paper offers a study on the Reed Xiaoli Anomaly (RXA) detection algorithm performance investigated for increasing number of spectral bands from 30, 50, 100 to all the spectral bands present in the HSI. The original RXA algorithm is formulated with Mahalanobis distance. In this study the RXA al is re-implemented with other different distance algorithms like Bhattacharya distance, Kullback-Leibler divergence, and Jeffery divergence and evaluated for any change in the performance. For the first part of investigation, the obtained results showed that the decreased number of spectral bands shows better performance in terms of receiver operating characteristic (ROC) obtained for cumulative probability values and false alarm rate. In the next part of study it is found that, the RXA algorithm with Jeffrey divergence has a comparable performance in ROC to that of the RX algorithm with Mahalanobis distance.

Support vector machine (SVM) is one of the most powerful technologies of machine learning, which has been widely concerned because of its remarkable performance. However, when dealing with the classification problem of large-scale datasets, the high complexity of SVM model leads to low efficiency and become impractical. Due to the sparsity of SVM in the sample space, this paper presents a new parallel data geometry analysis (PDGA) algorithm to reduce the training set of SVM, which helps to improve the efficiency of SVM training. The PDGA introduce Mahalanobis distance to measure the distance from each sample to its centroid. And based on this, proposes a method that can identify non support vectors and outliers at the same time to help remove redundant data. When the training set is further reduced, cosine angle distance analysis method is proposed to determine whether the samples are redundant data, ensure that the valuable data are not removed. Different from the previous data geometry analysis methods, the PDGA algorithm is implemented in parallel, which greatly saving the computational cost. Experimental results on artificial dataset and 6 real datasets show that the algorithm can adapt to different sample distributions. Which significantly reduce the training time and memory requirements without sacrificing the classification accuracy, and its performance is obviously better than the other five competitive algorithms.

Fast and accurate brushing is crucial in visual data exploration and sketch-based solutions are successful methods. In this article, we detail a solution, based on kernel density estimation, which computes a data subset selection in a scatterplot from a simple click-and-drag interaction. We explain how this technique relates to two alternative approaches, i.e., Mahalanobis brushing and CNN brushing. To study this relation, we conducted two user studies and present both a quantitative three-fold comparison as well as additional details about the prevalence of all possible cases in that each technique succeeds/fails. With this, we also provide a comparison between empirical modeling and implicit modeling by DL in terms of accuracy, efficiency, generality, and interpretability.

1. The quantification of Hutchison9s n-dimensional hypervolume has enabled substantial progress in community ecology, species niche analysis and beyond. While non-parametric methods for quantifying and comparing hypervolumes are popular, they do not support a partitioning of the different components and drivers of hypervolume variation. Here, we propose as alternative the use of multivariate normal distributions for the assessment and comparison of niche hypervolumes and introduce this as the multivariate-normal hypervolume (MVNH) framework. 2. The framework provides parametric measures of the size and dissimilarity of niche hypervolumes, each of which can be partitioned into biologically interpretable components. Specifically, We use 1) the determinant of the covariance matrix (i.e. the generalized variance) of a MVNH as a measure of total niche size, which can be partitioned into the components of univariate niche variances and a correlation component; and 2) the Bhattacharyya distance between two MVNHs as a measure of niche dissimilarity, which can be partitioned into the components of Mahalanobis distance between hypervolume centroids and the determinant ratio which measures hypervolume size difference. 3. We use empirical examples of community- and species-level analysis to demonstrate the new insights provided by these metrics. We show that the newly proposed framework enables us to quantify the relative contributions of different hypervolume components and to identify the drivers of functional diversity and environmental niche variation. 4. Our approach overcomes several operational and computational limitations of non-parametric methods and provides a framework that offers both unification and granularity in the assessment of niche volumes and differences, which has wide implications for understanding niche evolution, niche shifts and expansion during biotic invasions etc.

Purpose The purpose of this paper is to ensure the stable operation of satellites in orbit and to assist ground personnel in continuously monitoring the satellite telemetry data and finding anomalies in advance, which can improve the reliability of satellite operation and prevent catastrophic losses. Design/methodology/approach This paper proposes a deep auto-encoder (DAE) satellite anomaly advance warning framework for satellite telemetry data. Firstly, this study performs grey correlation analysis, extracts important feature attributes to construct feature vectors and builds the variational auto-encoder with bidirectional long short-term memory generative adversarial network discriminator (VAE/BLGAN). Then, the Mahalanobis distance is used to measure the reconstruction score of input and output. According to the periodic characteristic of satellite operation, a dynamic threshold method based on periodic time window is proposed. Satellite health monitoring and advance warning are achieved using reconstruction scores and dynamic thresholds. Findings Experiment results indicate DAE methods can probe that satellite telemetry data appear abnormal, trigger a warning before the anomaly occurring and thus allow enough time for troubleshooting. This paper further verifies that the proposed VAE/BLGAN model has stronger data learning ability than other two auto-encoder models and is sensitive to satellite monitoring data. Originality/value This paper provides a DAE framework to apply in the field of satellite health monitoring and anomaly advance warning. To the best of the authors’ knowledge, this is the first paper to combine DAE methods with satellite anomaly detection, which can promote the application of artificial intelligence in spacecraft health monitoring.

Balancing allocation of assigning units to two treatment groups to minimize the allocation differences is important in biomedical research. The complete randomization, rerandomization, and pairwise sequential randomization (PSR) procedures can be employed to balance the allocation. However, the first two do not allow a large number of covariates. In this article, we generalize the PSR procedure and propose a k-resolution sequential randomization (k-RSR) procedure by minimizing the Mahalanobis distance between both groups with equal group size. The proposed method can be used to achieve adequate balance and obtain a reasonable estimate of treatment effect. Compared to PSR, k-RSR is more likely to achieve the optimal value theoretically. Extensive simulation studies are conducted to show the superiorities of k-RSR and applications to the clinical synthetic data and GAW16 data further illustrate the methods.

Data in practice are often of high dimension and multivariate in nature. Detection of outliers has been one of the problems in multivariate analysis. Detecting outliers in multivariate data is difficult and it is not sufficient by using only graphical inspection. In this paper, a nontechnical and brief outlier detection method for multivariate data which are projection pursuit method, methods based on robust distance and cluster analysis are reviewed. The strengths and weaknesses of each method are briefly discussed.

Tightly-coupled GNSS/INS (Global Navigation Satellite System/Inertial Navigation System) integration is of importance to vehicle positioning. However, this integration technology has difficulty in achieving optimal positioning solutions for the dynamic systems involving strong nonlinearity and systematic modelling error. This paper proposes a new methodology to address the problem of tightly-coupled GNSS/INS integration. This methodology rigorously derives a novel adaptive CKF (Cubature Kalman Filter) with fading memory for kinematic modelling error and a new robust CKF with emerging memory for observation modelling error, using the concept of Mahalanobis distance without involving artificial empiricism. Based on this, a new CKF with both adaptability and robustness is further developed by fusing the results of the standard CKF, adaptive CKF and robust CKF via the principle of interacting multiple model (IMM). Simulation and experiment results together with comparison analysis prove that the proposed methodology can curb the interferences of both kinematic and observation modelling errors on state estimation, leading to improved positioning accuracy for vehicle positioning via tightly-coupled GNSS/INS integration.

There are several algorithms to solve many problems in grouping data. Grouping data is also known as clusterization, clustering takes advantage to solve some problems especially in business. In this note, we will modify the clustering algorithm based on distance principle which background of K-means algorithm (Euclidean distance). Manhattan, Mahalanobis-Euclidean, and Chebyshev distance will be used to modify the K-means algorithm. We compare the clustered result related to their accuracy, we got Mahalanobis - Euclidean distance gives the best accuracy on our experiment data, and some results are also given in this note.

The automatization of the Re-Identification of an individual across different video-surveillance cameras poses a significant challenge due to the presence of a vast number of potential candidates with a similar appearance. This task requires the learning of discriminative features from person images and a distance metric to properly compare them and decide whether they belong to the same person or not. Nevertheless, the fact of acquiring images of the same person from different, distant and non-overlapping views produces changes in illumination, perspective, background, resolution and scale between the person’s representations, resulting in appearance variations that hamper his/her re-identification. This article focuses the feature learning on automatically finding discriminative descriptors able to reflect the dissimilarities mainly due to the changes in actual people appearance, independently from the variations introduced by the acquisition point. With that purpose, such variations have been implicitly embedded by the Mahalanobis distance. This article presents a learning algorithm to jointly model features and the Mahalanobis distance through a Deep Neural Re-Identification model. The Mahalanobis distance learning has been implemented as a novel neural layer, forming part of a Triplet Learning model that has been evaluated over PRID2011 dataset, providing satisfactory results.

The empirical analysis of a typical gear fault diagnosis of five different classes has been studied in this article. The analysis was used to develop novel feature selection criteria that provide an optimum feature subset over feature ranking genetic algorithms for improving the planetary gear fault classification accuracy. We have considered traditional approach in the fault diagnosis, where the raw vibration signal was divided into fixed-length epochs, and statistical time-domain features have been extracted from the segmented signal to represent the data in a compact discriminative form. Scale-invariant Mahalanobis distance–based feature selection using ANOVA statistic test was used as a feature selection criterion to find out the optimum feature subset. The Support Vector Machine Multi-Class machine learning algorithm was used as a classification technique to diagnose the gear faults. It has been observed that the highest gear fault classification accuracy of 99.89% (load case) was achieved by using the proposed Mahalanobis-ANOVA Criterion for optimum feature subset selection followed by Support Vector Machine Multi-Class algorithm. It is also noted that the developed feature selection criterion is a data-driven model which will contemplate all the nonlinearity in a signal. The fault diagnosis consistency of the proposed Support Vector Machine Multi-Class learning algorithm was ensured through 100 Monte Carlo runs, and the diagnostic ability of the classifier has been represented using confusion matrix and receiver operating characteristics.

To avoid the decrease of system reliability due to insufficient component maintenance and the resource waste caused by excessive component maintenance, identifying the critical components of complex products is an effective way to improve the efficiency of maintenance activities. Existing studies on identifying critical components of complex products are mainly from two aspects i.e., topological properties and functional properties, respectively. In this paper, we combine these two aspects to establish a hybrid intuitionistic fuzzy set to incorporate the different types of attribute values. Considering the mutual correlation between attributes, a combination of AHP (Analytic Hierarchy Process) and Improved Mahalanobis-Taguchi System (MTS) is used to determine the λ-Shapley fuzzy measures for attributes. Then, the λ-Shapley Choquet integral intuitionistic fuzzy TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution) method is proposed for calculating the closeness degrees of components to generate their ranking order. Finally, a case study which is about the right gear of airbus 320 is taken as an example to verify the feasibility and effectiveness of this method. This novel methodology can effectively solve the critical components identification problem with different types of evaluation information and completely unknown weight information of attributes, which provides the implementation of protection measures for the system reliability of complex products.

Regional integrated energy system (RIES) provides a platform for coupling utilization of multi-energy and makes various energy demand from client possible. The suitable RIES composition scheme will upgrade energy structure and improve integrated energy utilization efficiency. Based on a RIES construction project in Jiangsu province, this paper proposes a new multi criteria decision-making (MCDM) method for the selection of RIES schemes. Because that subjective evaluation on RIES schemes benefit under criteria has uncertainty and hesitancy, intuitionistic trapezoidal fuzzy number (ITFN) which has the better capability to model ill-known quantities is presented. In consideration of risk attitude and interdependency of criteria, a new decision model with risk coefficients, Mahalanobis-Taguchi system and Choquet integral is proposed. Firstly, the decision matrices given by experts are normalized, and then are transformed to minimum expectation matrices according to different risk coefficients. Secondly, the weights of criteria from different experts are calculated by Mahalanobis-Taguchi system. Mobius transformation coefficients based on interaction degree are to calculate 2-order additive fuzzy measures, and then the comprehensive weights of criteria are obtained by fuzzy measures and Choquet integral. Thirdly, based on group decision consensus requirement, the weights of experts are obtained by the maximum entropy and grey correlation. Fourthly, the minimum expectation matrices are aggregated by the intuitionistic trapezoidal fuzzy Bonferroni mean operator. Thus, the ranking result according to the comparison rules using the minimum expectation and the maximum expectation is obtained. Finally, an illustrative example is taken in the present study to make the proposed method comprehensible.

This paper deals with the fault detection of a pneumatic control valve using canonical variate analysis (CVA). CVA can find the optimal linear combinations of p-window and f-window data, so that the correlation between these combinations can be maximized. Based on CVA, the p-window data is considered by traditional hotelling T2 statistic and squared prediction error (SPE) indicators, the corresponding fault detection rates (FDR) are low. In order to improve the FDR, a detection indicator based on SMD (square of the Mahalanobis distance) of the residual is proposed in this paper. The proposed indicator considers not only the information in the p-window data, but also that of the f-window data, which can improve the FDRs. The proposed techniques have been validated using a Development and Application of Methods for Actuator Diagnosis in Industrial Control Systems (DAMADICS) benchmark. It concludes that 14 out of the 19 faults can be successfully detected using the proposed method (CVA-SMD). Simulation results have shown that the CVA-SMD can improve the FDR compared with existing CVA-T2 and CVA-SPE methods. Experiments based on real-world data have also demonstrated that the CVA-SMD has better performance than existing PCA-T2, PCA-SPE, PCA-SMD, CVA-T2 and CVA-SPE methods.

In this work, we deal with a robust fitting of a wrapped normal model to multivariate circular data. Robust estimation is supposed to mitigate the adverse effects of outliers on inference. Furthermore, the use of a proper robust method leads to the definition of effective outlier detection rules. Robust fitting is achieved by a suitable modification of a classification-expectation-maximization algorithm that has been developed to perform a maximum likelihood estimation of the parameters of a multivariate wrapped normal distribution. The modification concerns the use of complete-data estimating equations that involve a set of data dependent weights aimed to downweight the effect of possible outliers. Several robust techniques are considered to define weights. The finite sample behavior of the resulting proposed methods is investigated by some numerical studies and real data examples.

Blind deconvolution (BD) methods have proven to be effective tools for condition monitoring of gearbox. Nevertheless, due to the severe operating environment and complex structure in the wind turbine (WT) gearbox, the prior knowledge of fault period is hard to obtain accurately, which results in a challenge to the traditional BD algorithms that exceedingly relies on this information. Motivated by this limitation, a novel BD approach based on sparse subspace recoding (SSRBD) is proposed for the condition monitoring of WT gearbox. In this work, singular value decomposition is initially introduced to convert the raw signal from the input space to feature subspaces. The coefficient of variation is then constructed to guide the choice of inverse filter length. Successively, an iterative Mahalanobis distance is designed to cluster the sensitive subspace with rich fault information. Finally, build upon the robust principal component analysis, the objective features are further separated by means of sparse recoding. The effectiveness and robustness of the proposed SSRBD are validated through several comparative analyses and experimental cases. The consequences demonstrate that the proposed approach overcomes the dependence of prior information and domain knowledge, while extracts the fault feature more effectively than the state-of-the-art methods.

Unmanaged cell inconsistency may cause accelerated battery degradation or even thermal runaway accidents in electric vehicles (EVs). Accurate cell inconsistency evaluation is a prerequisite for efficient battery health management to maintain safe and reliable operation and is also vital for battery second-life utilization. This article presents a cell inconsistency evaluation model for series-connected battery systems based on real-world EV operation data. The open-circuit voltage (OCV), internal resistance, and charging voltage curve are extracted as consistency indicators (CIs) from a large volume of electric taxis’ operation data. The Thevenin equivalent circuit model is adopted to delineate battery dynamics, and an adaptive forgetting factor recursive least-squares method is proposed to reduce the fluctuation phenomenon in model parameter identification. With a modified robust regression method, the evolution characteristics of the three CIs are analyzed. The Mahalanobis distance in combination with the density-based spatial clustering of applications with noise is employed to comprehensively evaluate the multiparameter inconsistency state of a battery system based on the CIs. The results show that the proposed method can effectively assess cell inconsistency with high robustness and is competent for real-world applications.

A small physical change in the eye influences the entire neural information process along the visual pathway, causing perceptual errors and behavioral changes. Astigmatism, a refractive error in which visual images do not evenly focus on the retina, modulates visual perception, and the accompanying neural processes in the brain. However, studies on the neural representation of visual stimuli in astigmatism are scarce. We investigated the relationship between retinal input distortions and neural bias in astigmatism and how modulated neural information causes a perceptual error. We induced astigmatism by placing a cylindrical lens on the dominant eye of human participants, while they reported the orientations of the presented Gabor patches. The simultaneously recorded electroencephalogram activity revealed that stimulus orientation information estimated from the multivariate electroencephalogram activity was biased away from the neural representation of the astigmatic axis and predictive of behavioral bias. The representational neural dynamics underlying the perceptual error revealed the temporal state transition; it was transiently dynamic and unstable (approximately 350 ms from stimulus onset) that soon stabilized. The biased stimulus orientation information represented by the spatially distributed electroencephalogram activity mediated the distorted retinal images and biased orientation perception in induced astigmatism., In this study, we investigated the neural sources of perceptual error induced by the astigmatic distortion of the retinal image. We found a systematic distortion in the neural orientation representation estimated from the multivariate analysis of the electroencephalogram recordings, which supports the observed patterns of perceptual error.

The Mahalanobis-Taguchi system (MTS) is a relatively new multi-dimensional pattern recognition technology that combines Mahalanobis distance (MD) with Taguchi's robust engineering for diagnosis and...

The integration of Global Navigation Satellite System (GNSS) with Inertial Navigation Systems (INS) has been actively researched and widely applied as it can provide reliable positioning information continuously. In recent years, Micro Electro Mechanical Systems (MEMS) technology achieves rapid development and Micro Electro Mechanical Systems and Inertial Measurement Unit (MEMS-IMU) has aroused wide concern due to its excellent properties in some cases. However, the observations from MEMS-IMU are easy to be influenced by motion state and location environment because of its manufacturing process. It is not easy to judge whether gross errors are in the state model or the observation model by the widely adopted robust filter based on innovation. In this contribution, we present an improved robust filter with a double state model on the basis of the chi-square distribution of the square of the Mahalanobis distance. The vehicle motion model acts as the external constraint information and can be adopted to construct robust statistic with the results from INS mechanization. And then a robust factor was determined to adjust the observation noise covariance matrix. To evaluate the performance of this method, the simulation test and the field test based on locomotive platform of Nottingham Geospatial Institute (NGI) were carried out. According to the results, in the simulation test, the position improvements are 33%, 30% in the north and east directions; in the real test, the loosely and tightly coupled was adopted and the position accuracy can be improved by about 50–60% in the horizontal direction and the improvement of the pitch and the roll accuracy was lower than the azimuth accuracy due to poor observability and experimental scene which is of the characteristics of small elevation change. Therefore, the proposed robust filter could diminish the effect of the gross error from MEMS-IMU and enhance the integrated system.

Vehicle control problems have been historically using simple control structures with lookup tables that were designed to use vehicle steady-state characterization to calibrate and are suited better for older systems with lower complexity/dimension. However, the increasing complexity of the vehicle systems and stringent regulations requires optimal control of the dynamic vehicle systems. In this brief, we present an efficient online identification methodology with a composite local model structure that will be critical to the new control paradigm. We first introduce the concept of evolving spatial–temporal filters (STFs) that dynamically transform an incoming input–output data stream into a nonlinear combination of local models. The local models are weighed by an array of weights corresponding to the compatibility of the input–output data to a set of ellipsoidal clusters that partition the input–output space. The filters exploit ellipsoidal-shape evolving clusters as function bases and a distance metric defined as a combination of Mahalanobis distance and scaled local model prediction error. Parameters of the filters and local models can be updated simultaneously online, making adaptive optimization a possibility for vehicle systems. Evolving clustering and recursive least square techniques are exploited to simultaneously update the cluster parameters and local linear models. We apply the developed algorithm to the modeling of a turbocharged internal combustion engine that is a highly nonlinear and complex system. Promising performance is demonstrated both in a high-fidelity simulator and on an experimental vehicle.

Untracked space debris is the principal threat to operational satellites’ functioning whose services have become a fundamental part of our daily life. Though some specialised sensors can detect objects down to sub-cm sizes in geostationary Earth orbit, only objects larger than 30 cm are currently being catalogued. Thus, small debris are only seldom observed, typically for a short amount of time when surveying the sky. Having to deal with short arcs, the data association’s problem becomes relevant: one must find more observations of the same resident space object to precisely determine its orbit. This paper develops a new method enabled by differential algebra for track initialisation and catalogue build-up, within the framework of multi-target tracking. This is compared to literature methods that build on the concept of the admissible region and attributable to solve the problem of correlating sparse observations. The comparison is carried out on synthetic measurements and real optical observations obtained by the ZimSMART telescope on consecutive nights. Furthermore, simulated observations are used to assess whether raw data in the tracklets can be exploited to reduce the admissible region’s size. Though the gain in computational efficiency is only limited, this paper effectively shows an alternative method to the Mahalanobis distance, where the success of correlation is less affected by the time separation of two observations.

In this article, we propose new waveform design procedures and classification schemes to improve target classification performance nonadaptively in radar systems. The new designs and schemes are all inspired by 2-class and multiclass Fisher discriminant analysis. The proposed system does not require as much computational capability as adaptive waveform design systems while also overcoming: 1) angular uncertainty in classifying high fidelity targets and 2) drops in performance experienced by nonadaptive systems when classification is extended to more than two targets. The waveform design procedure is based on an optimization problem to find the waveform that maximizes the objective function inspired by Fisher analysis under constant energy constraint. We also derive two closed-form solutions for the optimization problems under certain conditions for the 2-class and multiclass cases. All the methods are tested using synthetic and real data to show the performance of the proposed methods against the average Mahalanobis distance (AMD) nonadaptive waveform design and classifier.

The fault diagnosis of a rotating multi-blade system has been conducted using the vibration characteristics of the system. For the fault diagnosis, multi-dimensional features related to the system vibration characteristics were extracted and used for monitoring the system health condition in most of previous studies. Recently, blade tip timing (BTT) method becomes increasingly popular to measure the vibration signals of rotating multi-blade systems. Due to the under-sampling characteristics of the method, relatively low frequency components of the vibration signal can be only collected with the BTT method. In this study, a statistical index called the Mahalanobis distance is defined and employed for the fault diagnosis of a rotating multi-blade system having a crack. The effects of crack existence and signal to noise ratio on the reliability of the proposed method using BTT signals obtained with a simulation model are investigated in this study.

Electrical and electronic industry is one of Malaysia’s leading industries which covers around 24.5% in manufacturing production sector. With a continuous innovation of the Industry, inductor component gets higher demand from customer and it is good if there is a study to convince that those factors are really significant to the production as well. Meanwhile, the current costing being used is difficult to access the complete activities required for each workstation and need separate analysis to measure the un-used capacity in term of resources and cost. The objective of this work is to clarify the relationship between Mahalanobis-Taguchi system (MTS) and time driven activity-based costing (TDABC) in the electronic industry. The data collection is focused on inductor component by consiedring the historical data in 2018. MTS is used as a method to optimize various parameters while TDABC is used to measure the un-used capacity by constructing the time equation and capacity cost rate. There are 7 parameters considered which are condition of wire, condition of winding, condition of epoxy, condition of core, condition of lead part, condition of marking and condition of soldering. As a result, MTS is successfully developed the normal and abnormal Mahalanobis distance (MD). In February, the normal MD is 0.9998 and the abnormal is 15.6538 with 2 significant parameters with signal to noise is 0.1244. In addition, there are 3 parameters consistently influenced along 10 months such as condition of core, condition of lead part and condition of soldering and 2 parameters are not consistently influenced such as condition of epoxy and condition marking. On the other hand, the total used and un-used capacity of time are 257124.02 minutes and 5217031.43 minutes respectively while the total of used and un-used of cost are MYR6,296,493.10 and MYR6214807.07 respectively. Eventually, this work concludes that both methods are a great tool and feasible to be implemented in the electronic industry.

Great efforts are now underway to control the coronavirus 2019 disease (COVID-19). Millions of people are medically examined, and their data keep piling up awaiting classification. The data are typically both incomplete and heterogeneous which hampers classical classification algorithms. Some researchers have recently modified the popular KNN algorithm as a solution, where they handle incompleteness by imputation and heterogeneity by converting categorical data into numbers. In this article, we introduce a novel KNN variant (KNNV) algorithm that provides better results as demonstrated by thorough experimental work. We employ rough set theoretic techniques to handle both incompleteness and heterogeneity, as well as to find an ideal value for K. The KNNV algorithm takes an incomplete, heterogeneous dataset, containing medical records of people, and identifies those cases with COVID-19. We use in the process two popular distance metrics, Euclidean and Mahalanobis, in an effort to widen the operational scope. The KNNV algorithm is implemented and tested on a real dataset from the Italian Society of Medical and Interventional Radiology. The experimental results show that it can efficiently and accurately classify COVID-19 cases. It is also compared to three KNN derivatives. The comparison results show that it greatly outperforms all its competitors in terms of four metrics: precision, recall, accuracy, and F-Score. The algorithm given in this article can be easily applied to classify other diseases. Moreover, its methodology can be further extended to do general classification tasks outside the medical field.

In a number of recent articles Riani, Cerioli, Atkinson and others advocate the technique of monitoring robust estimates computed over a range of key parameter values (Cerioli et al., 2018; Riani et al., 2019). Through this approach the diagnostic tools of choice can be tuned in such a way that highly robust estimators which are as efficient as possible are obtained. This approach is applicable to different robust multivariate estimates like S- and MM-estimates, MVE and MCD as well as to the Forward Search in which monitoring is part of the robust method. Key tool for detection of multivariate outliers and for monitoring of robust estimates are the scaled Mahalanobis distances and statistics related to these distances. However, the results obtained with this tool in case of compositional data might be unrealistic since compositional data contain relative rather than absolute information and need to be transformed to the usual Euclidean geometry before the standard statistical tools can be applied. Several specific transformations have been introduced, but Filzmoser and Hron (2008) show that the transformation with the best properties with respect to robust estimates and keeping invariant the Mahalanobis distances is the ilr (isometric log-ratio) transformation. To illustrate the problem of monitoring compositional data and to demonstrate the usefulness of monitoring in this case we start with a simple example and then analyze a real life data set presenting the technological structure of manufactured exports which, as an indicator of their quality, is an important criterion for understanding the relative position of countries measured by their industrial competitiveness. The analysis is conducted with the R package fsdaR, which makes the analytical and graphical tools provided in the MATLAB FSDA library available for R users.

Rolling bearings are one of the most vulnerable parts in rotating machines. This paper presents a novel approach to identify the rolling bearings fault based on variational mode decomposition (VMD) and Mahalanobis distance support vector machine (MDSVM). In this work, since the original vibration signal contains a lot of noise, we use wavelet threshold method to denoise the original vibration signal. The vibration signals are generally non-linear, to extract feature, VMD has been employed to reconstruct signals. When raw signals are decomposed by VMD, according to the center frequency of each decomposed mode, the number of modes is selected. Then we calculate the sample entropy of the decomposed modal component, which is considered as the feature and input of support vector machine (SVM). The Euclidean distance is usually used in the calculation of the Gaussian kernel function of the SVM, which cannot measure the distance between two samples accurately, so we combine the Mahalanobis distance with SVM, construct a Gaussian function kernel based on Mahalanobis distance, and propose a classifier model based on Mahalanobis distance Gaussian function kernel. The model integrates the parameter solutions of the Mahalanobis distance function and the support vector machine into the same framework, which makes full use of the advantages of both and makes it easier to get the solution of the parameters. Finally, all feature vectors are utilized to train improved SVM, with which the fault modes of rolling bearings are identified. The experimental results show that the proposed method has better diagnosing performance.