Discriminative Approach Research Articles

Combination of spectral and textural features of hyperspectral imaging for the authentication of the diet supplied to fattening cattle

This study explored the potential of hyperspectral imaging in the near infrared region (NIR-HSI) as a non-destructive and rapid tool to discriminate among two beef fattening diets. For that purpose, a feeding trial was carried out with a total of 24 purebred Pirenaica calves. Twelve of them were fed barley and straw (BS) while 11 animals were finished on vegetable by-products (VBPR). When comparing the reference measurements of the meat coming from those animals, only the total collagen ratio expressed the feeding effect (p-value<0.05). To undertake the authentication procedure, two discrimination approaches were run: partial least squares discriminant analysis (PLS-DA) and radial basis function-support vector machine (RBF-SVM). To precisely extract spectral and textural information from the lean portion of the meat steaks, various techniques were executed, such as principal component (PC) images, competitive adaptive reweighted sampling (CARS) for selecting optimal wavelengths, and gray-level-co-occurrence matrix (GLCM). After hyperspectral imaging and the combination of their own texture features, samples were classified according to feeding diet with an overall accuracy of 72.92% for PLS-DA and 80.56% for RBF-SVM. So, the potential of using HSI technology to authenticate the meat obtained from beef supplied a diet based on circular economy techniques was made in evidence.

Advancements in food authentication using soft independent modelling of class analogy (SIMCA): a review

Abstract Food authentication verifies the match between product characteristics and claims and it is crucial in a globalized and complex food sector. Currently, class-modelling approaches, such as soft independent modelling of class analogy (SIMCA), are powerful tools for assessing food authenticity. The aim of this review is to discuss the application of SIMCA for food authentication and to describe the conceptual differences between discriminant and class-modelling approaches. The discussion of research articles is organized around three elements: (i) the research objectives, (ii) the analytical methodologies, and (iii) the food products investigated. Moreover, the challenges and future perspectives considering the development of innovative food products are discussed. Adulteration is the most investigated food authentication issue, followed by verification of geographical origin. Food authenticity appeared to be predominantly evaluated using non-destructive spectroscopy. Overall, the articles collectively cover a broad spectrum of food categories, representing those most prone to adulteration. However, there is a notable lack of food authentication studies on innovative food products, underscoring the urgency for further research in this field.

Mining Text Value in Japanese Translation Teaching Based on KH Coder Text Data

Abstract Vocabulary serves as the cornerstone of Japanese translation instruction, with the identification of high-frequency terms within educational texts yielding significant pedagogical insights. This study employs the KH-Coder software to extract feature words frequently used in Japanese translation teaching materials. Through a combination of covariate network analysis, sentiment analysis, and hierarchical clustering, the paper delves into the intrinsic textual value of these educational resources. Furthermore, it integrates both generative and discriminative approaches to conduct dependency parsing of Japanese translation. By incorporating probabilistic functions, the methodology enhances contextual understanding, ensuring that each annotation unit within a sentence is optimally tagged. Analysis of the CGTT-35 corpus through KH-Coder reveals a list of the top 50 high-frequency words in Japanese translation teaching, each appearing over 100 times. This enables students to use these frequent words and sentence structures as cues to master common phrases and familiarize themselves with core content. The approach has proven effective in enhancing students’ receptivity to Japanese translation instruction and their comprehension capabilities, offering valuable implications for teaching strategies.

An energy-aware heart disease prediction system using ESMO and optimal deep learning model for healthcare monitoring in IoT

ABSTRACT: The Internet of Things (IoT), which provides seamless connectivity between people and things, improves our quality of life. In the medical field, predictive analytics can help transform a reactive healthcare (HC) strategy into a proactive one. The HC industry embraces cutting-edge artificial intelligence and machine learning (ML) technologies. ML’s area of deep learning has the revolutionary potential to reliably analyze massive volumes of data quickly, produce insightful revelations and solve challenging issues. This article proposes an energy-aware heart disease prediction (HDP) system based on enhanced spider monkey optimization (ESMO) and a weight-optimized neural network for an IoT-based HC environment. The proposed work consists of two essential phases: energy-efficient data transmission and HDP. In energy-efficient transmission, the cluster leaders are optimally selected using ESMO and the cluster formation is done based on Euclidean distance. In HDP, the patient data are collected from the dataset, and essential features are extracted. After that, the dimensionality reduction is carried out using the modified linear discriminant analysis approach to reduce over-fitting issues. Finally, the HDP uses the enhanced Archimedes weight-optimized deep neural network (EAWO-DNN). The simulation findings demonstrate that the proposed optimal clustering mechanism enhances the network’s lifespan by consuming minimal energy compared to the existing techniques. Also, the proposed EAWO-DNN classifier achieves higher prediction accuracy, precision, recall and f-measure than the conventional methods for predicting heart disease in IoT. Communicated by Ramaswamy H. Sarma

Optical Solitons for the Dispersive Concatenation Model with Power-Law Nonlinearity by the Complete Discriminant Approach

The present paper is a sequel to the paper on the topic with Kerr law nonlinearity. For power-law nonlinearity, it was derived through balancing principle that the solitons would exist for two values of the power-law parameter, namely n = 1 or n = 2. The current paper derives dark and singular solitons for the latter value of n since the first case was already covered in a previous report that was dedicated to address the model with Kerr law.

A data extension framework of seismic-induced gravelly soil liquefaction based on semi-supervised methods

In studies of gravelly soil liquefaction discrimination, the number of training samples significantly affects model performance. However, historical earthquakes provide limited instances of gravelly soil liquefaction, resulting in no guarantee of the generalization performance of existing models. Additionally, valuable data from liquefiable critical layer points and unlabeled points outside the critical layer are often overlooked in stratigraphic exploration, despite their high value. Therefore, in this paper, a data extension method for semi-supervised learning based on the Markov chain Monte Carlo–Bayesian network (MCMC-BN) method is proposed and a semi-MCMC-BN gravelly soil liquefaction discrimination model is constructed. By effectively mining the available site liquefaction information, the proposed method increases the sample size of the gravelly soil liquefaction dataset by approximately 4 times, expanding 120 initial liquefaction samples to 476. The results demonstrate that the extended liquefaction database improves the distribution of liquefaction data for gravelly soils. The semi-supervised learning model outperforms the supervised learning model, highlighting the feasibility of the semi-supervised approach for gravelly soil liquefaction discrimination. In addition, debris flow data for the Wenchuan area, China, are used to validate the effectiveness of the proposed method. Finally, the impacts of the confidence level magnitude, critical soil layer selection, and different machine learning methods on the data extension outcomes are discussed. The expanded database can provide data support to studying gravelly soil liquefaction, and the proposed data expansion framework can be used for related research in other fields.

Zeroth- and first-order difference discrimination for unsupervised domain adaptation

Unsupervised domain adaptation transfers empirical knowledge from a label-rich source domain to a fully unlabeled target domain with a different distribution. A core idea of many existing approaches is to reduce the distribution divergence between domains. However, they focused only on part of the discrimination, which can be categorized into optimizing the following four objectives: reducing the intraclass distance between domains, enlarging the interclass distances between domains, reducing the intraclass distances within domains, and enlarging the interclass distances within domains. Moreover, because few methods consider multiple types of objectives, the consistency of data representations produced by different types of objectives has not yet been studied. In this paper, to address the above issues, we propose a zeroth- and first-order difference discrimination (ZFOD) approach for unsupervised domain adaptation. It first optimizes the above four objectives simultaneously. To improve the discrimination consistency of the data across the two domains, we propose a first-order difference constraint to align the interclass differences across domains. Because the proposed method needs pseudolabels for the target domain, we adopt a recent pseudolabel generation method to alleviate the negative impact of imprecise pseudolabels. We conducted an extensive comparison with nine representative conventional methods and seven remarkable deep learning-based methods on four benchmark datasets. Experimental results demonstrate that the proposed method, as a conventional approach, not only significantly outperforms the nine conventional comparison methods but is also competitive with the seven deep learning-based comparison methods. In particular, our method achieves an accuracy of 93.4% on the Office+Caltech10 dataset, which outperforms the other comparison methods. An ablation study further demonstrates the effectiveness of the proposed constraint in aligning the objectives.

Deep learning approach for discrimination of liver lesions using nine time-phase images of contrast-enhanced ultrasound.

Contrast-enhanced ultrasound (CEUS) shows different enhancement patterns depending on the time after administration of the contrast agent. The aim of this study was to evaluate the diagnostic performance of liver nodule characterization using our proposed deep learning model with input of nine CEUS images. A total of 181 liver lesions (48 benign, 78 hepatocellular carcinoma (HCC), and 55 non-HCC malignant) were included in this prospective study. CEUS were performed using the contrast agent Sonazoid, and in addition to B-mode images before injection, image clips were stored every minute up to 10min. A deep learning model was developed by arranging three ResNet50 transfer learning models in parallel. This proposed model allowed inputting up to nine datasets of different phases of CEUS and performing image augmentation of nine images synchronously. Using the results, the correct prediction rate, sensitivity, and specificity between "benign" and "malignant" cases were analyzed for each combination of the time phase. These accuracy values were also compared with the washout score judged by a human. The proposed model showed performance superior to the referential standard model when the dataset from B-mode to the 10-min images were used (sensitivity: 93.2%, specificity: 65.3%, average correct answer rate: 60.1%). It also maintained 90.2% sensitivity and 61.2% specificity even when the dataset was limited to 2min after injection, and this accuracy was equivalent to or better than human scoring by experts. Our proposed model has the potential to identify tumor types earlier than the Kupffer phase, but at the same time, machine learning confirmed that Kupffer-phase Sonazoid images contain essential information for the classification of liver nodules.

Supervised and unsupervised machine learning approaches for prediction and geographical discrimination of Iranian saffron ecotypes based on flower-related and phytochemical attributes

A two-year field experiment (2014–2016; Zanjan, Iran) was conducted to monitor potential diversity pattern and adaptability power among 18 Iranian saffron ecotypes under Zanjan climatological conditions using seven flower-related and three qualitative traits (crocin, picrocrocin, and safranal, determined by UV–visible spectra), and analyzed by supervised and unsupervised approaches. A range of variability was recorded among the ecotypes, and despite some exceptions, overall, saffron corms produced higher amounts of studied features across the second year. The Feizabad ecotype was recommended to acquire maximum qualitative criteria (category I; based on ISO Normative 3632 grading system), while for flower-related parameters several ecotypes (e.g., Ghaien, Bardeskan, Torbat-Jam, and Gonabad) could be applied for Zanjan climatological conditions. Based on the results of Leave-One-Out Cross-Validation (LOOCV), various prediction values were computed for all 10 classifiers of LDA, QDA, FDA, MDA, RDA, Naive Bayes, Decision Tree, Linear SVM, Radial SVM, and Random Forest in terms of accuracy, sensitivity and specificity parameters. Among which, Random Forest and LDA with the values of 0.91 and 0.78 possessed the highest and the lowest amounts of accuracy, respectively. Finally, considering the highest accuracy value of the superior classification model of Random forest, both feature subsets of “FFW, FDW, Picrocrocin, Safranal, and Crocin” and “SFW, FDW, Picrocrocin, Safranal, and Crocin” were nominated as the most powerful elements (comparing to the remaining 1021 feature subsets) to make accurate discrimination between Khorasan and non-Khorasan saffron ecotypes. The results, overall, revealed that saffron ecotypes followed different responses under Zanjan climatological circumstances, and Random Forest is more suitable for accurately predicting saffron corms from different provenances.

Microbiome and plant cell transformation trigger insect gall induction in cassava.

Several specialised insects can manipulate normal plant development to induce a highly organised structure known as a gall, which represents one of the most complex interactions between insects and plants. Thus far, the mechanism for insect-induced plant galls has remained elusive. To study the induction mechanism of insect galls, we selected the gall induced by Iatrophobia brasiliensis (Diptera: Cecidomyiidae) in cassava (Euphorbiaceae: Manihot esculenta Crantz) as our model. PCR-based molecular markers and deep metagenomic sequencing data were employed to analyse the gall microbiome and to test the hypothesis that gall cells are genetically transformed by insect vectored bacteria. A shotgun sequencing discrimination approach was implemented to selectively discriminate between foreign DNA and the reference host plant genome. Several known candidate insertion sequences were identified, the most significant being DNA sequences found in bacterial genes related to the transcription regulatory factor CadR, cadmium-transporting ATPase encoded by the cadA gene, nitrate transport permease protein (nrtB gene), and arsenical pump ATPase (arsA gene). In addition, a DNA fragment associated with ubiquitin-like gene E2 was identified as a potential accessory genetic element involved in gall induction mechanism. Furthermore, our results suggest that the increased quality and rapid development of gall tissue are mostly driven by microbiome enrichment and the acquisition of critical endophytes. An initial gall-like structure was experimentally obtained in M. esculenta cultured tissues through inoculation assays using a Rhodococcus bacterial strain that originated from the inducing insect, which we related to the gall induction process. We provide evidence that the modification of the endophytic microbiome and the genetic transformation of plant cells in M. esculenta are two essential requirements for insect-induced gall formation. Based on these findings and having observed the same potential DNA marker in galls from other plant species (ubiquitin-like gene E2), we speculate that bacterially mediated genetic transformation of plant cells may represent a more widespread gall induction mechanism found in nature.

Optical solitons for the concatenation model with differential group delay by the complete discriminant approach

Optical solitons for the concatenation model with differential group delay by the complete discriminant approach

Attention driven CWT-deep learning approach for discrimination of Radar PRI modulation

With the proliferation of radio frequency (RF) systems and radar applications, Electronic Warfare (EW) is receiving increasing importance. The analysis of the radar signals is a critical EW task that decides the nature of counter employments. In an Electronic Support (ES) system, the challenge is to detect hostile radiation sources efficiently and trigger a counter response. Detection of types of Pulse Repetition Interval (PRI) modulation of radar signal significantly facilitates the manifestation of RF emitters during recognition which is difficult in a dense EW environment. Recent developments in artificial intelligence (AI) methods suggest that this emerging technology can be effective for such purposes. In this direction, an automatic approach for recognizing several kinds of complex PRI modulation based on Continuous Wavelet Transform (CWT) and a combination of the vanilla Convolutional Neural Network (CNN), a multi-head self-attention (MHSA) mechanism and the popular Long Short-Term Memory (LSTM) is proposed. The CWT is used to decompose the PRI modulation sequence and obtain different time–frequency components. Further, aided by the proposed CNN-MHSA-LSTM combination, the features extracted from the CWT 2D-scalograms are used to execute PRI modulation discrimination. In this method, the vanilla CNN is employed for the extraction of deep features to figure out the class details while capturing the spatial attributes. Thereafter, to improve the discriminative power of the entire framework a MHSA mechanism is used. The temporal attributes are acquired by the LSTM which works in concert with the CNN for executing the detection of the PRI classes based on the extracted features. Also to assess the effectiveness of the proposed method, three models based on ResNet, popular CNN and SqueezeNet are implemented for benchmark comparison in terms of overall performance and complexity. The simulation results show that the proposed method enhances performance and achieves robustness in the noise-filled and imperfect channel knowledge environment. The best recognition accuracy is 98.3% with 50% spurious pulses in the environment which fluctuates with imperfect channel knowledge cases.

Multimodal Kernel-based discriminant correlation analysis data-fusion approach: an automated autism spectrum disorder diagnostic system.

Autism spectrum disorder (ASD) diagnostic systems, based on association of multimodal tools such as combination of Electroencephalogram (EEG) and eye-tracking, have emerged as an analytical to provide objective biomarkers. However, the existing feature-redundancy-based systems have lacked in providing knowledge of fusion approaches and robust feature-set. The present paper aims to reduce disorder homogeneity by proposing a multimodal diagnostic system which can incorporate multimodal data. The paper has collected simultaneous-data from three modalities (laptop-performance tool, EEG machine, and Eye-tracker) fused the recorded computational, neural and visual data. The multimodal features are analyzed via proposed multimodal Kernel-based discriminant correlation analysis (MKDCA) fusion approach and classified using state-of-the-art machine-learning classifiers. The proposed framework has considered the distinct cardinality of the feature vectors and fused the group structure among multiple samples after ranking them in increasing order. As per the results, the proposed multimodal system provided fused feature set of 11 influential features out of total 39 features. The SVM classifier has diagnosed ASD with 92% testing accuracy and 0.988 AUC(ROC). The proposed automated fusion-based system has the potential to classify disorder by reducing the disorder heterogeneity and stratifying ASD individuals into homogeneous sub-groups. In future, the correlation of reduced feature set with ASD clinical symptoms accounted by screening scales can provide clinical relevance of proposed model.

Discriminative Approach Lung Diseases and COVID-19 from Chest X-Ray Images Using Convolutional Neural Networks: A Promising Approach for Accurate Diagnosis

Medical imaging treatment is one of the best-known computer science disciplines. It can be used to detect the presence of several diseases such as skin cancer and brain tumors, and since the arrival of the coronavirus (COVID-19), this technique has been used to alleviate the heavy burden placed on all health institutions and personnel, given the high rate of spread of this virus in the population. One of the problems encountered in diagnosing people suspected of having contracted COVID-19 is the difficulty of distinguishing symptoms due to this virus from those of other diseases such as influenza, as they are similar. This paper proposes a new approach to distinguishing between lung diseases and COVID-19 by analyzing chest x-ray images using a convolutional neural network (CNN) architecture. To achieve this, pre-processing was carried out on the dataset using histogram equalization, and then we trained two sub-datasets from the dataset using the Train et Test, the first to be used in the training phase and the second to be used in the model validation phase. Then a CNN architecture composed of several convolution layers and fully connected layers was deployed to train our model. Finally, we evaluated our model using two different metrics: the confusion matrix and the receiver operating characteristic. The simulation results recorded are satisfactory, with an accuracy rate of 96.27%.

Performance Profiling in Handball Using Discriminative Variables and its Practical Applications

Performance profiles of teams performance highlight areas of weaknesses and strengths for coaches to inform their decision-making on how to spend their limited training time with athletes. This study used a stepwise discriminative analysis approach comparing one successful team’s (TEAM) performances through five consecutive seasons against a) other top four teams (TOP4) and b) teams with a final rank between 5th and eight (LOW) in a semi-professional league. The predictive model created was used to set forth a performance profile for the selected team. A total of 95 matches of the TEAM’s matches from the last five seasons are in the analysis. The objective was to create a performance profile with relevant performance indicators selected based on the discriminant analysis results of the selected TEAM and discuss its practical applicability. For matches against other TOP4 teams, the predictive model created consisted of three variables; legal stops, blocked shots and 9 m shots, classifying 72.6% correctly. The LOW ranked teams model had six variables and correctly classified 94.4% of cases (assists, blocked shots, legal stops, the goalkeeper saved shots, 2-minute exclusion, and shot efficiency). The selected variables are presented in Table 4, with medians and a 95% confidence interval of the median as a team performance profile. The profile provides the coaches with two models containing values that can serve as a reference for this team’s performance. The profile of this TEAM’s performances during the last five seasons generally aligns with the variables associated with success in other studies in female handball.

Source Discrimination of Mine Water by Applying the Multilayer Perceptron Neural Network (MLP) Method—A Case Study in the Pingdingshan Coalfield

In a complex multiaquifer mine, discriminant approaches that have previously been presented cannot successfully locate water sources. With multiple processing layers, computing models may learn representations of data at various levels of abstraction. Modern technology in many domains has significantly enhanced these models. The problem of distinguishing the source of mine water in mines and tunnels has been addressed by studying the hydrochemical components of the Pingdingshan coalfield and applying the multilayer perceptron neural network (MLP) method to discriminate the source of the mine water. There were five types of mine water in the Pingdingshan coalfield. Each type of water was encoded with the numbers 0 to 4. The one-hot code method was used to encode the numbers, which is the output set. On the basis of hydrochemical data processing, the MLP model was developed using a characteristic ion contrast on aquifers with distinctive chemical properties. The research results show that two hidden layers (with 10 neurons in each hidden layer) for the model were fit for completing the prediction process with a better performance. This approach enabled us to discriminate water sources for the Pingdingshan coalfield and could be tried for other coalfields with similar hydrogeological conditions.

Differentiating True and False Cinnamon: Exploring Multiple Approaches for Discrimination.

This study presents a comprehensive literature review that investigates the distinctions between true and false cinnamon. Given the intricate compositions of essential oils (EOs), various discrimination approaches were explored to ensure quality, safety, and authenticity, thereby establishing consumer confidence. Through the utilization of physical-chemical and instrumental analyses, the purity of EOs was evaluated via qualitative and quantitative assessments, enabling the identification of constituents or compounds within the oils. Consequently, a diverse array of techniques has been documented, encompassing organoleptic, physical, chemical, and instrumental methodologies, such as spectroscopic and chromatographic methods. Electronic noses (e-noses) exhibit significant potential for identifying cinnamon adulteration, presenting a rapid, non-destructive, and cost-effective approach. Leveraging their capability to detect and analyze volatile organic compound (VOC) profiles, e-noses can contribute to ensuring authenticity and quality in the food and fragrance industries. Continued research and development efforts in this domain will assuredly augment the capacities of this promising avenue, which is the utilization of Artificial Intelligence (AI) and Machine Learning (ML) algorithms in conjunction with spectroscopic data to combat cinnamon adulteration.

Image Segmentation of the Sudd Wetlands in South Sudan for Environmental Analytics by GRASS GIS Scripts

This paper presents the object detection algorithms GRASS GIS applied for Landsat 8-9 OLI/TIRS data. The study area includes the Sudd wetlands located in South Sudan. This study describes a programming method for the automated processing of satellite images for environmental analytics, applying the scripting algorithms of GRASS GIS. This study documents how the land cover changed and developed over time in South Sudan with varying climate and environmental settings, indicating the variations in landscape patterns. A set of modules was used to process satellite images by scripting language. It streamlines the geospatial processing tasks. The functionality of the modules of GRASS GIS to image processing is called within scripts as subprocesses which automate operations. The cutting-edge tools of GRASS GIS present a cost-effective solution to remote sensing data modelling and analysis. This is based on the discrimination of the spectral reflectance of pixels on the raster scenes. Scripting algorithms of remote sensing data processing based on the GRASS GIS syntax are run from the terminal, enabling to pass commands to the module. This ensures the automation and high speed of image processing. The algorithm challenge is that landscape patterns differ substantially, and there are nonlinear dynamics in land cover types due to environmental factors and climate effects. Time series analysis of several multispectral images demonstrated changes in land cover types over the study area of the Sudd, South Sudan affected by environmental degradation of landscapes. The map is generated for each Landsat image from 2015 to 2023 using 481 maximum-likelihood discriminant analysis approaches of classification. The methodology includes image segmentation by ‘i.segment’ module, image clustering and classification by ‘i.cluster’ and ‘i.maxlike’ modules, accuracy assessment by ‘r.kappa’ module, and computing NDVI and cartographic mapping implemented using GRASS GIS. The benefits of object detection techniques for image analysis are demonstrated with the reported effects of various threshold levels of segmentation. The segmentation was performed 371 times with 90% of the threshold and minsize = 5; the process was converged in 37 to 41 iterations. The following segments are defined for images: 4515 for 2015, 4813 for 2016, 4114 for 2017, 5090 for 2018, 6021 for 2019, 3187 for 2020, 2445 for 2022, and 5181 for 2023. The percent convergence is 98% for the processed images. Detecting variations in land cover patterns is possible using spaceborne datasets and advanced applications of scripting algorithms. The implications of cartographic approach for environmental landscape analysis are discussed. The algorithm for image processing is based on a set of GRASS GIS wrapper functions for automated image classification.

Machine-Learning Aided NO2 Discrimination with an Array of Graphene Chemiresistors Covalently Functionalized by Diazonium Chemistry.

Boosted by the emerging need for highly integrated gas sensors in IoT ecosystems, electronic noses (e-noses) are gaining interest for the detection of specific molecules over a background of interfering gases. The sensing of nitrogen dioxide (NO2) is particularly relevant for applications in environmental monitoring and precision medicine. Here we present an easy and efficient functionalization procedure to covalently modify graphene layers, taking advantage of diazonium chemistry. Separate graphene layers were functionalised with one of three different aryl rings 4-nitrophenyl, 4-carboxyphenyl, and 4-bromophenyl). The distinct mod(ified graphene layers were assembled with a pristine layer into an e-nose for NO2 discrimination. A remarkable sensitivity to NO2 is demonstrated through exposures to gaseous solutions with NO2 concentrations in the 1-10 ppm range at room temperature. Then, the discrimination capability of the sensor array is tested by carrying out exposures to several interfering gases and analyzing the data through multivariate statistical analysis. This analysis shows that the e-nose can discriminate NO2 among all the interfering gases in a two-dimensional principal component analysis space. Finally, the e-nose is trained to accurately recognize NO2 contributions with a linear discriminant analysis approach, thus providing a metric for discrimination assessment with a prediction accuracy above 95%.

Methodological approach for the automatic discrimination of pictorial materials using fused hyperspectral imaging data from the visible to mid-infrared range coupled with machine learning methods

Hyperspectral imaging represents a powerful tool for the study of artwork’s materials since it permits to obtain simultaneously information about the spectral behavior of the materials and their spatial distribution. By combining hyperspectral images performed on several spectral intervals (visible, near infrared and mid-infrared ranges) through chemometric methods it is possible to clearly identify most of the materials used in painting (i.e., pigments, dyes, varnishes, and binders). Moreover, in the last decade, the development of machine learning algorithms coupled with comprehensive and continuously updated databases opens new perspective on the automatic recognition of pictorial materials.In this work, we propose a novel procedure to support the automatic discrimination of pictorial materials consisting in a mid-level data fusion on imaging datasets coming from two commercial hyperspectral cameras, in the 400–1000 nm and 1000–2500 nm spectral ranges, respectively, and a MAcroscopic Fourier Transform InfRared scanning in reflection mode (MA-rFTIR), in the 7000 to 350 cm-1 (1428 nm − 28 μm) spectral range. The automatic recognition of 102 pictorial mock-ups from the fused data is performed by testing the performance of ECOC-SVM (error-correcting output coding and support vector machine) model obtaining a good predictive result with only few pixels that are confused with other classes. The methodology described in this paper demonstrates that an accurate paint layer multiclass recognition is feasible, and the use of chemometric approaches solves some challenges involving the study of materials.