FIGURE 10 - uploaded by José Manuel Amigo
Content may be subject to copyright.
Upper: RGB image of cheese sample with plastic contaminant. Average AE standard deviation of ROI spectra for cheese and plastic are shown on the up right hand side of the upper portion of the figure. Middle: Average NIR hyperspectral image of the cheese/plastic dataset. Lower: results of building LDA and PLS-DA classifiers and applying it to the hypercube are shown. All of the pixels corresponding to the plastic contaminant are predicted as belonging to class '1' and all of the pixels corresponding to the cheese are predicted as belonging to class '0'. Also, the result of applying PLSR is shown, by applying a threshold of 0.5.
Contexts in source publication
Context 1
... this example, a slice of cheese with plastic contamination was imaged using the NIR HSI system described in [75]. An RGB image of the cheese and contamination is shown in Figure 10. The region imaged by the NIR HSI system is delineated by a red rectangle. ...
Context 2
... or cheese) for clas- sification modelling. Plastic spectra were assigned a categorical variable of 1, while cheese spectra were assigned a value of 0. The mean spectrum of each object is also shown in Figure 10, top right. It is clear that the spectral profile of the contaminant is very different to that of the cheese sample, indicating that this will not be a difficult classification problem. ...
Context 3
... distance is typically used to measure the distance to class mean. In Figure 10 (lower section), the results of applying Fishers LDA to the cheese/plastic dataset are shown. The LDA classifier works perfectly for dis- criminating between these two objects, due to the significant spectral Lower: results of building LDA and PLS-DA classifiers and applying it to the hypercube are shown. ...
Context 4
... shown in this figure is the mean NIR HSI image, from which the plastic is clearly identifiable as distinct from the cheese background. This is mainly due to differences in reflected light inten- sity from each object (the plastic reflects almost twice the amount of light that the cheese does, as can be seen in the spectra shown in Figure 10). The advan- tage of applying LDA in this case is that the classifier is based on the spectral profile of each class, rather than average light intensity, which could vary due to extraneous factors, such as sample height and morphology. ...
Context 5
... of the PLSR model to hypercube data results in a predicted image of spatially varying intensity for each class. However, thresholding this image so that all pixel values above 0.5 were classified as plastic and all pixel values less than 0.5 were classified as black results in a classification image identical to the LDA classification image (Figure 10). ...
Similar publications
The increasingly common application of the near-infrared (NIR) hyperspectral imaging technique to the analysis of food powders has led to the need for optical characterization of samples. This study was aimed at exploring the feasibility of quantifying penetration depth of NIR hyperspectral imaging light for milk powder. Hyperspectral NIR reflectan...
Citations
... Principal Component Analysis (PCA) [1][2][3] is one the most fundamental tools in chemometrics for data compression and visualization of complex data sets. It is widely employed as an exploratory tool for all kinds of data, such as in hyperspectral imaging [4,5], quality control [6], complex mixtures [7], as a basic tool for classification methods [8] or in process analytical technology [9,10], and for an endless number of different applications. ...
... Unsupervised principal component analysis (PCA) is a versatile method adopting to process spectral data which gives an overview of complex multivariable data [31]. PCA reduces the data dimensionality and orthogonalizes the multi-dimensional spectra into few PC scores replacing original variables to improve the speed in modeling calibration and prediction. ...
Consumer trust in fruit industry is partly depended on the labeling of its geographical origin. The accurate label is not only of great significance to consumers and retailers, but also for their brands shaping. This work evaluated the utility of Fourier transform near-infrared spectroscopy (FT-NIRS) in identifying the geographical origins of Chinese pears including Laiyang pear (Pyrus bretschneideri cv. Laiyang), Dangshan pear (Pyrus bretschneideri Rehd. cv. Dangshansuli), and Korla pear (Pyrus sinkiangensis T. T. Yu). FT-NIR spectra of 316 pears from three different geographical origins were obtained in the spectral range of 10,000–4000 cm⁻¹. Preliminary investigation was first conducted based on principal component analysis (PCA) to confirm the similar spectral features of samples within one class, and the results showed that PC1 and PC2 were effective for the identification. Partial least square discriminant analysis (PLS-DA) models based on full spectra indicated that the optimal classification based on second-order derivative (Der2) preprocessed spectra achieved 100% correct classification rates (CCRs) in all datasets. After that, the multivariate selection strategies including two-dimensional correlation spectroscopy (2D-COS), competitive adaptive reweighted sampling (CARS) combined with successive projection algorithm (SPA), and uninformative variable elimination (UVE) combined with SPA were individually optimized for selecting the characteristic wavenumbers from full FT-NIR spectra. On the basis of simplified PLS-DA models establishment, the CCRs in calibration, cross-validation, and prediction sets of the selected CARS-SPA-PLS-DA model achieved 98.58%, 98.11%, and 98.08%, respectively. This research demonstrated that FT-NIRS combined with multivariate analysis can be successfully employed to identify the geographical origins of Chinese pears.
... This technique is widely used in the industrial field, in the agri-food and in the recycling sector for quality control and process monitoring [23,24]. HSI, along with chemometrics, combines the benefits of conventional spectroscopy with those of classical image analysis and has been applied by different authors to study plastic degradation [25] and classify plastics [26,27]. ...
Plastic in agriculture is frequently used to protect crops and its use boosts output, enhances food quality, contributes to minimize water consumption, and reduces the environmental impacts of agricultural activities. On the other hand, end-of-life plastic management and disposal are the main issues related to their presence in this kind of environment, especially in respect of plastic degradation, if not properly handled (i.e., storage places directly in contact with the ground, exposure of stocks to meteoric agents for long periods, incorrect or incomplete removal). In this study, the possibility of using an in situ near infrared (NIR: 1000–1700 nm) hyperspectral imaging detection architecture for the recognition of various plastic wastes in agricultural soils in order to identify their presence and also assess their degradation from a recovery/recycling perspective was explored. In more detail, a Partial Least Squares—Discriminant Analysis (PLS-DA) classifier capable of identifying plastic waste from soil was developed, implemented, and set up. Results showed that hyperspectral imaging, in combination with chemometric approaches, allows the utilization of a rapid, non-destructive, and non-invasive analytical approach for characterizing the plastic waste produced in agriculture, as well as the potential assessment of their lifespan.
... The combination of image techniques with vibrational spectroscopy has also determined the possibility to spatially analyse the sample in addition to collect its spectrum (Amigo et al., 2013;Cozzolino & Roberts, 2016;Manley, 2014;Li et al., 2014). For instance, in the implementation of hyperspectral image analyses, this method determines that an image is constructed by a hypercube with a three-dimensional dataset consisting of one spectral and two spatial dimensions (Amigo et al., 2013;Cozzolino & Roberts, 2016;Manley, 2014;Li et al., 2014). ...
... The combination of image techniques with vibrational spectroscopy has also determined the possibility to spatially analyse the sample in addition to collect its spectrum (Amigo et al., 2013;Cozzolino & Roberts, 2016;Manley, 2014;Li et al., 2014). For instance, in the implementation of hyperspectral image analyses, this method determines that an image is constructed by a hypercube with a three-dimensional dataset consisting of one spectral and two spatial dimensions (Amigo et al., 2013;Cozzolino & Roberts, 2016;Manley, 2014;Li et al., 2014). This technique has allowed for the analysis of images that can be sensitive to detailed changes in the concentration and distribution of chemicals, enabling the visualization of the distribution of metabolites or compounds in the sample matrix (Amigo et al., 2013;Cozzolino & Roberts, 2016;Manley, 2014;Li et al., 2014;Saha & Manickavasagan, 2021;Rahaman et al., 2019;Virlet et al., 2017). ...
... For instance, in the implementation of hyperspectral image analyses, this method determines that an image is constructed by a hypercube with a three-dimensional dataset consisting of one spectral and two spatial dimensions (Amigo et al., 2013;Cozzolino & Roberts, 2016;Manley, 2014;Li et al., 2014). This technique has allowed for the analysis of images that can be sensitive to detailed changes in the concentration and distribution of chemicals, enabling the visualization of the distribution of metabolites or compounds in the sample matrix (Amigo et al., 2013;Cozzolino & Roberts, 2016;Manley, 2014;Li et al., 2014;Saha & Manickavasagan, 2021;Rahaman et al., 2019;Virlet et al., 2017). Furthermore, the utilization of hyperspectral image analysis has boosted our capability for a better understanding of the properties of the samples (Amigo et al., 2013;Cozzolino & Roberts, 2016;Manley, 2014;Saha & Manickavasagan, 2021;Rahaman et al., 2019;Virlet et al., 2017). ...
... Indeed, in hyperspectral images, there can be found thousands of spectra for the same measured variable, and a spectrum that represents the sample is calculated. The mean spectra for each sample are combined to create the data set (X), which is used with regression techniques to predict the corresponding observed values (y) [27]. For these purposes, methods like multiple linear regression (MLR), principal component regression (PCR), partial least square regression (PLSR), support vector machine regression (SVR), or artificial neural networks (ANNs) are broadly used [28]. ...
During the last two decades, hyperspectral imaging (HSI) has been one of the most studied and applied techniques in the field of nondestructive monitoring systems for the fruit and vegetable supply chain. This review provides HSI technical aspects (i.e., device features) and data analysis approaches (i.e., data processing and qualitative/quantitative modeling) for fresh-cut products, focusing on the different applications which the literature offers and the possible scale-up for process monitoring. Moreover, new frontiers in the development of possible process analytical technologies of cost-effective and hand-held HSI devices are presented and discussed. Even though the performance of these new proximal sensing tools needs to be carefully evaluated, new applicative research perspectives in the development of a proximal sensing approach based on HSI sensor networks are ready to be studied and developed for finding field applications (i.e., precision agriculture, food processing, and more) and enabling faster and more convenient analysis while maintaining the accuracy and capabilities of traditional HSI systems.
... The best pre-processing strategies and combinations for each rule were chosen starting from the maximum cluster separation detected by PCA. The different pre-processing combinations were used for the following purposes: Standard Normal Variate (SNV) and Multiplicative Scatter Correction (MSC) were applied to reduce the effects of light scattering (Rinnan et al., 2009); Savitzky-Golay (SG) Derivative and Smoothing were used to emphasize detected spectral differences and to reduce noise in the analyzed samples (Rinnan et al., 2009;Amigo et al., 2013); finally, Mean Center (MC) was applied for data centering (Vidal and Amigo, 2012). ...
... PCA was chosen to reveal relations between variables (i.e., wavelengths) and samples (i.e., plastic flakes), detecting outliers, finding and quantifying patterns and generating new hypotheses. PCA allows the processed spectral data to be decomposed into several Principal Components (PCs), linear combinations of the original spectral data, embedding the spectral variations of each collected spectral dataset (Amigo et al., 2013;Bro and Smilde, 2014). ...
... An extension to PCA known as principal component regression (PCR) uses the PCs of the explanatory variables to perform standard regression to describe a dependent variable. Application of PCA includes the study of the relation between dough properties and baking quality [89], characterization of desiccation of wheat kernels [55], and as a dimensionality reduction technique in many other applications [90]. ...
Grains intended for human consumption or feedstock are typically high-value commodities that are marketed based on either their visual characteristics or compositional properties. The combination of visual traits, chemical composition and contaminants is generally referred to as grain quality. Currently, the market value of grain is quantified at the point of receival, using trading standards defined in terms of visual criteria of the bulk grain and chemical constituency. The risk for the grower is that grain prices can fluctuate throughout the year depending on world production, quality variation and market needs. The assessment of grain quality and market value on-farm, rather than post-farm gate, may identify high- and low-quality grain and inform a fair price for growers. The economic benefits include delivering grain that meets specifications maximizing the aggregate price, increasing traceability across the supply chain from grower to consumer and identifying greater suitability of differentiated products for high-value niche markets, such as high protein product ideal for plant-based proteins. This review focuses on developments that quantify grain quality with a range of spectral sensors in an on-farm setting. If the application of sensor technologies were expanded and adopted on-farm, growers could identify the impact and manage the harvesting operation to meet a range of quality targets and provide an economic advantage to the farming enterprise.
... According to Amigo et al. [84], support vector machine, stepwise multiple linear regression, partial least squares regression, and artificial neural networks (ANNs) are the most commonly used chemometric and preprocessing techniques in spectroscopy. These chemometric methods in combination with IR spectroscopy have been used to develop bioreceptor-free sensors and methods for various applications, including the detection of microorganisms, their metabolites, spoilage, and their diseases, as evidenced in some recent studies by Spyrelli et al. [85], Cebrian et al. [86], Alexandrakis et al. [87], and Azadshahraki et al. [88], respectively. ...
Microbial pathogens cause a quarter of all deaths worldwide annually due to deadly infectious diseases. Nevertheless, the fast and precise identification of pathogens remains one of the most challenging tasks in the medical sector. Early identification and characterization of microbes through medical diagnosis could pave the way for specific treatment strategies that could dramatically improve infection management, reduce healthcare costs, mitigate increasing antimicrobial resistance, and save numerous lives. To date, numerous traditional and molecular methods have been employed to diagnose illnesses with proven accuracy, reliability, and efficiency. Here, we have reviewed the most reliable tools that are prerequisites for the rapid detection of microbes. In particular, the remarkable roles of surface-enhanced Raman sca ering, Fourier-transform infrared, electrochemical impedance, near-infrared, and MALDI-TOF/TOF in the identification and characterization of pathogenic microbes are discussed in detail. The approaches described herein cover broad ranges of biomedical applications, including the diagnosis of clinical infectious diseases, epidemiology, detection of vector-borne diseases, food security, phytosanitary monitoring, biosensing, and food-and waterborne pathogen detection. Considering the current pandemic outbreak, this review briefly emphasizes the importance of rapid detection and upgraded tools for early diagnosis to prevent the loss of lives.
... More in detail, the spectra preprocessing was performed as follows: raw spectra were preliminary cut, at the beginning and at the end of the investigated wavelength range, in order to eliminate unwanted effects due to lighting/background noise. The different preprocessing combinations were used for the following purposes: detrend was applied to reduce the effects of light scattering [23]; Savitzky-Golay (SG) derivative was used to emphasise detected spectral differences and to reduce noise in the analysed samples [23,25]; finally, Mean Centre (MC) was applied for data centring [26]. ...
The aim of this study was the development of a new multi-analytical approach to evaluate chemical alterations and differences in the element content in relation to arsenic (As) in the As hyperaccumulator fern P. vittata. P. vittata plants were grown on two natural As-rich soils with either high or moderate As (750 and 58 mg/kg). Dried samples from plant tissues were then analysed by means of micro-energy dispersive X-ray fluorescence spectrometry (μ-XRF), attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) and hyperspectral imaging (HSI) with a multivariate approach. The As and micro- and macronutrients content was evaluated by μ-XRF and a significant correlation between As, potassium (K), iron (Fe), calcium (Ca) and manganese (Mn) contents were found at both moderate and high As levels. The same samples were then analysed by ATR-FTIR spectroscopy and HSI (SWIR range, 1000–2500 nm). Interestingly, by FTIR analysis it was found that the main differences between the control and the As-contaminated samples are due to the intensity of the absorption band related to polysaccharides (i.e., cellulose, hemicellulose and pectin), lignin, lipid and amide groups. The same chemical alterations were detected by an HSI analysis and all the FTIR and HSI data were validated by a PCA analysis. These results suggest a possible complexation of As ions with the amide group. Moreover, the proposed μ-XRF, HSI and ATR-FTIR combining approach could be a promising strategy to monitor in-field phytoremediation approaches by directly controlling the As content in plants.
... The resulting output from the use of this technology are georeferenced image cubes referred to as hypercubes ( Figure 3). A 3-D image hypercube consists of a two-dimensions X * Y (2-D) narrowband image array, organised along the spectral (λ) axis [61]. Hyperspectral sensors can detect numerous closely spaced wavelength ranges. ...
The potential of precision viticulture has been highlighted since the first studies performed in the context of viticulture, but especially in the last decade there have been excellent results have been achieved in terms of innovation and simple application. The deployment of new sensors for vineyard monitoring is set to increase in the coming years, enabling large amounts of information to be obtained. However, the large number of sensors developed and the great amount of data that can be collected are not always easy to manage, as it requires cross-sectoral expertise. The preliminary section of the review presents the scenario of precision viticulture, highlighting its potential and possible applications. This review illustrates the types of sensors and their operating principles. Remote platforms such as satellites, unmanned aerial vehicles (UAV) and proximal platforms are also presented. Some supervised and unsupervised algorithms used for object-based image segmentation and classification (OBIA) are then discussed, as well as a description of some vegetation indices (VI) used in viticulture. Photogrammetric algorithms for 3D canopy modelling using dense point clouds are illustrated. Finally, some machine learning and deep learning algorithms are illustrated for processing and interpreting big data to understand the vineyard agronomic and physiological status. This review shows that to perform accurate vineyard surveys and evaluations, it is important to select the appropriate sensor or platform, so the algorithms used in post-processing depend on the type of data collected. Several aspects discussed are fundamental to the understanding and implementation of vineyard variability monitoring techniques. However, it is evident that in the future, artificial intelligence and new equipment will become increasingly relevant for the detection and management of spatial variability through an autonomous approach.
