Journal of Spectral Imaging JSI—Journal of Spectral Imaging is an online, Open Access, peer-review journal publishing high-quality papers in the rapidly growing field of spectral, hyperspectral and chemical imaging. https://www.impopen.com/jsi-toc/7 IM Publications Open en © IM Publications Open LLP Journal of Spectral Imaging 2040-4565 © IM Publications Open LLP info@impopen.com Near infrared hyperspectral images and pattern recognition techniques used to identify etiological agents of cotton anthracnose and ramulosis https://doi.org/10.1255/jsi.2018.a8 Hyperspectral imaging near infrared (HSI-NIR) has the potential to be used as a non-destructive approach for the analysis of new microbiological matrices of agriculture interest. This article describes a new method for accurately and rapidly classifying the etiological agents Colletotrichum gossypii (CG) and C. gossypii var. cephalosporioides (CGC) grown in a culture medium, using scattering reflectance HSI-NIR and multivariate pattern recognition analysis. Five strains of CG and 46 strains of CGC were used. CG and CGC strains were grown on Czapek-agar medium at 25?°C under a 12-hour photoperiod for 15?days. Molecular identification was performed as a reference for the CG and CGC classes by polymerase chain reaction of the intergenic spacer region of rDNA. The scattering coefficient µs and the absorption coefficient µa were obtained, which resulted in a µs value for CG of 1.37 × 1019 and for CGC of 5.83 × 10–11. These results showed that the use of the standard normal variate was no longer essential and reduced the spectral range from 1000–2500?nm to 1000–1381?nm. The results evidenced two type II errors for the CG 457-2 and CGC 39 samples in the soft independent modelling model of the analogy model. There were no classification errors using the algorithm of the successive projections for variable selection in linear discriminant analysis (SPA-LDA). A parallel validation of the results obtained with SPA-LDA was performed using a box plot analysis with the 11 variables selected by SPA, in which there were no outliers for the HSI-NIR models. The new HSI-NIR and SPA-LDA procedures for the classification of CG and CGC etiological agents are noted for their greater analytical speed, accuracy, simplicity, lower cost and non-destructive nature. Near infrared hyperspectral images and pattern recognition techniques used to identify etiological agents of cotton anthracnose and ramulosis

J. Spectral Imaging 7, a8 (2018). doi:10.1255/jsi.2018.a8

Hyperspectral imaging near infrared (HSI-NIR) has the potential to be used as a non-destructive approach for the analysis of new microbiological matrices of agriculture interest. This article describes a new method for accurately and rapidly classifying the etiological agents Colletotrichum gossypii (CG) and C. gossypii var. cephalosporioides (CGC) grown in a culture medium, using scattering reflectance HSI-NIR and multivariate pattern recognition analysis. Five strains of CG and 46 strains of CGC were used. CG and CGC strains were grown on Czapek-agar medium at 25?°C under a 12-hour photoperiod for 15?days. Molecular identification was performed as a reference for the CG and CGC classes by polymerase chain reaction of the intergenic spacer region of rDNA. The scattering coefficient µs and the absorption coefficient µa were obtained, which resulted in a µs value for CG of 1.37 × 1019 and for CGC of 5.83 × 10–11. These results showed that the use of the standard normal variate was no longer essential and reduced the spectral range from 1000–2500?nm to 1000–1381?nm. The results evidenced two type II errors for the CG 457-2 and CGC 39 samples in the soft independent modelling model of the analogy model. There were no classification errors using the algorithm of the successive projections for variable selection in linear discriminant analysis (SPA-LDA). A parallel validation of the results obtained with SPA-LDA was performed using a box plot analysis with the 11 variables selected by SPA, in which there were no outliers for the HSI-NIR models. The new HSI-NIR and SPA-LDA procedures for the classification of CG and CGC etiological agents are noted for their greater analytical speed, accuracy, simplicity, lower cost and non-destructive nature.

]]>
Near infrared hyperspectral images and pattern recognition techniques used to identify etiological agents of cotton anthracnose and ramulosis Priscila S.R. Aires Francisco F. Gambarra-Neto Wirton M. Coutinho Alderi E. Araujo Gilvan Ferreira da Silva Josivanda P.G. Gouveia Everaldo P. Medeiros doi:10.1255/jsi.2018.a8 J. Spectral Imaging 7, a8 (2018) 2018-04-24 Journal of Spectral Imaging 2018-04-24 7 1 17 10.1255/jsi.2018.a8 https://doi.org/10.1255/jsi.2018.a8
Semi-supervised learning of hyperspectral image segmentation applied to vine tomatoes and table grapes https://doi.org/10.1255/jsi.2018.a7 Nowadays, quality inspection of fruit and vegetables is typically accomplished through visual inspection. Automation of this inspection is desirable to make it more objective. For this, hyperspectral imaging has been identified as a promising technique. When the field of view includes multiple objects, hypercubes should be segmented to assign individual pixels to different objects. Unsupervised and supervised methods have been proposed. While the latter are labour intensive as they require masking of the training images, the former are too computationally intensive for in-line use and may provide different results for different hypercubes. Therefore, a semi-supervised method is proposed to train a computationally efficient segmentation algorithm with minimal human interaction. As a first step, an unsupervised classification model is used to cluster spectra in similar groups. In the second step, a pixel selection algorithm applied to the output of the unsupervised classification is used to build a supervised model which is fast enough for in-line use. To evaluate this approach, it is applied to hypercubes of vine tomatoes and table grapes. After first derivative spectral preprocessing to remove intensity variation due to curvature and gloss effects, the unsupervised models segmented 86.11% of the vine tomato images correctly. Considering overall accuracy, sensitivity, specificity and time needed to segment one hypercube, partial least squares discriminant analysis (PLS-DA) was found to be the best choice for in-line use, when using one training image. By adding a second image, the segmentation results improved considerably, yielding an overall accuracy of 96.95% for segmentation of vine tomatoes and 98.52% for segmentation of table grapes, demonstrating the added value of the learning phase in the algorithm. Semi-supervised learning of hyperspectral image segmentation applied to vine tomatoes and table grapes

J. Spectral Imaging 7, a7 (2018). doi:10.1255/jsi.2018.a7

Nowadays, quality inspection of fruit and vegetables is typically accomplished through visual inspection. Automation of this inspection is desirable to make it more objective. For this, hyperspectral imaging has been identified as a promising technique. When the field of view includes multiple objects, hypercubes should be segmented to assign individual pixels to different objects. Unsupervised and supervised methods have been proposed. While the latter are labour intensive as they require masking of the training images, the former are too computationally intensive for in-line use and may provide different results for different hypercubes. Therefore, a semi-supervised method is proposed to train a computationally efficient segmentation algorithm with minimal human interaction. As a first step, an unsupervised classification model is used to cluster spectra in similar groups. In the second step, a pixel selection algorithm applied to the output of the unsupervised classification is used to build a supervised model which is fast enough for in-line use. To evaluate this approach, it is applied to hypercubes of vine tomatoes and table grapes. After first derivative spectral preprocessing to remove intensity variation due to curvature and gloss effects, the unsupervised models segmented 86.11% of the vine tomato images correctly. Considering overall accuracy, sensitivity, specificity and time needed to segment one hypercube, partial least squares discriminant analysis (PLS-DA) was found to be the best choice for in-line use, when using one training image. By adding a second image, the segmentation results improved considerably, yielding an overall accuracy of 96.95% for segmentation of vine tomatoes and 98.52% for segmentation of table grapes, demonstrating the added value of the learning phase in the algorithm.

]]>
Semi-supervised learning of hyperspectral image segmentation applied to vine tomatoes and table grapes Jeroen van Roy Niels Wouters Bart De Ketelaere Wouter Saeys doi:10.1255/jsi.2018.a7 J. Spectral Imaging 7, a7 (2018) 2018-03-06 Journal of Spectral Imaging 2018-03-06 7 1 18 10.1255/jsi.2018.a7 https://doi.org/10.1255/jsi.2018.a7
Unsupervised classification of individual foodborne bacteria from a mixture of bacteria cultures within a hyperspectral microscope image https://doi.org/10.1255/jsi.2018.a6 Salmonella is a leading cause of foodborne illness. Traditional detection methods require lengthy incubation periods or expensive reagent kits. Hyperspectral microscope images (HMIs) have been previously investigated as a method for early and rapid detection of bacteria by using a spectral signature that is unique to the organism. Previous HMI use with bacteria has consisted of supervised classification with hypercubes collected for single culture images isolated from highly selective growth media. In order to move forward with HMI as a detection tool in the food industry, unsupervised classification of bacteria cells in mixed culture HMIs was investigated. Four foodborne bacteria cultures, S. Typhimurium (ST) E. coli (Ec), S. aureus (Sa) and L. innocua (Li) were combined in seven different culture combinations with HMIs collected between 450 nm and 800 nm. A k-means divisive cluster analysis (CA) was implemented and mixed culture image sets were found to contain between two and four clusters. CA cluster accuracy was obtained by assigning a dummy variable of the proposed CA classification, then carrying out a discriminant analysis. From the mixed culture HMIs, 700 bacteria cells were classified and accuracies were between 91.92% and 100%, with six of the seven HMI sets resulting in >?97% accuracies. A distance measure between clusters was applied to identify unknown clusters based on single culture reference samples of the four bacteria used. Results showed that the CA has potential for unsupervised classification of bacteria cells, but the distance metric was not an adequate method for identifying the unknown cluster based on reference spectra, potentially due to the collinearity amongst bacteria spectra. Unsupervised classification of individual foodborne bacteria from a mixture of bacteria cultures within a hyperspectral microscope image

J. Spectral Imaging 7, a6 (2018). doi:10.1255/jsi.2018.a6

Salmonella is a leading cause of foodborne illness. Traditional detection methods require lengthy incubation periods or expensive reagent kits. Hyperspectral microscope images (HMIs) have been previously investigated as a method for early and rapid detection of bacteria by using a spectral signature that is unique to the organism. Previous HMI use with bacteria has consisted of supervised classification with hypercubes collected for single culture images isolated from highly selective growth media. In order to move forward with HMI as a detection tool in the food industry, unsupervised classification of bacteria cells in mixed culture HMIs was investigated. Four foodborne bacteria cultures, S. Typhimurium (ST) E. coli (Ec), S. aureus (Sa) and L. innocua (Li) were combined in seven different culture combinations with HMIs collected between 450 nm and 800 nm. A k-means divisive cluster analysis (CA) was implemented and mixed culture image sets were found to contain between two and four clusters. CA cluster accuracy was obtained by assigning a dummy variable of the proposed CA classification, then carrying out a discriminant analysis. From the mixed culture HMIs, 700 bacteria cells were classified and accuracies were between 91.92% and 100%, with six of the seven HMI sets resulting in >?97% accuracies. A distance measure between clusters was applied to identify unknown clusters based on single culture reference samples of the four bacteria used. Results showed that the CA has potential for unsupervised classification of bacteria cells, but the distance metric was not an adequate method for identifying the unknown cluster based on reference spectra, potentially due to the collinearity amongst bacteria spectra.

]]>
Unsupervised classification of individual foodborne bacteria from a mixture of bacteria cultures within a hyperspectral microscope image Matthew Eady Bosoon Park doi:10.1255/jsi.2018.a6 J. Spectral Imaging 7, a6 (2018) 2018-03-06 Journal of Spectral Imaging 2018-03-06 7 1 13 10.1255/jsi.2018.a6 https://doi.org/10.1255/jsi.2018.a6
Concentration monitoring with near infrared chemical imaging in a tableting press https://doi.org/10.1255/jsi.2018.a5 Monitoring powder potency and homogeneity is important in achieving real-time release testing in a continuous tablet manufacturing operation. If quality related issues are encountered, monitoring powder potency inside a feed frame offers a last opportunity to intervene in the process before tablet compression. Feed frame monitoring methods based on near infrared (NIR) spectroscopy have been increasingly reported in recent years. New process analytical tools with the potential of being deployed alone or in combination with NIR spectroscopy for feed frame monitoring are now available commercially. The present study evaluated the potential of near infrared chemical imaging (NIR CI) for in-line monitoring of a prototype pharmaceutical composition containing ascorbic acid (AA), microcrystalline cellulose and dicalcium phosphate. NIR spectroscopy was the reference method. In-line calibration models based on partial least square regression were developed and validated with a range of AA concentrations. The ability of NIR spectroscopy and NIR CI to predict concentrations in test runs was ascertained both independently and in combination. NIR CI, with a single bandpass filter, predicted AA concentrations—present at commercially relevant concentrations—with acceptable accuracy. Comparative results showed that NIR CI has the potential for in-line monitoring of blend concentrations inside feed frames. In addition to the advantage of increased sample size, it also has the potential to detect segregation inside feed frames. Concentration monitoring with near infrared chemical imaging in a tableting press

J. Spectral Imaging 7, a5 (2018). doi:10.1255/jsi.2018.a5

Monitoring powder potency and homogeneity is important in achieving real-time release testing in a continuous tablet manufacturing operation. If quality related issues are encountered, monitoring powder potency inside a feed frame offers a last opportunity to intervene in the process before tablet compression. Feed frame monitoring methods based on near infrared (NIR) spectroscopy have been increasingly reported in recent years. New process analytical tools with the potential of being deployed alone or in combination with NIR spectroscopy for feed frame monitoring are now available commercially. The present study evaluated the potential of near infrared chemical imaging (NIR CI) for in-line monitoring of a prototype pharmaceutical composition containing ascorbic acid (AA), microcrystalline cellulose and dicalcium phosphate. NIR spectroscopy was the reference method. In-line calibration models based on partial least square regression were developed and validated with a range of AA concentrations. The ability of NIR spectroscopy and NIR CI to predict concentrations in test runs was ascertained both independently and in combination. NIR CI, with a single bandpass filter, predicted AA concentrations—present at commercially relevant concentrations—with acceptable accuracy. Comparative results showed that NIR CI has the potential for in-line monitoring of blend concentrations inside feed frames. In addition to the advantage of increased sample size, it also has the potential to detect segregation inside feed frames.

]]>
Concentration monitoring with near infrared chemical imaging in a tableting press Himmat Dalvi Clémence Fauteux-Lefebvre Jean- Maxime Guay Nicolas Abatzoglou Ryan Gosselin doi:10.1255/jsi.2018.a5 J. Spectral Imaging 7, a5 (2018) 2018-03-05 Journal of Spectral Imaging 2018-03-05 7 1 18 10.1255/jsi.2018.a5 https://doi.org/10.1255/jsi.2018.a5
Classification in hyperspectral images by independent component analysis, segmented cross-validation and uncertainty estimates https://doi.org/10.1255/jsi.2018.a4 Independent component analysis combined with various strategies for cross-validation, uncertainty estimates by jack-knifing and critical Hotelling’s T2 limits estimation, proposed in this paper, is used for classification purposes in hyperspectral images. To the best of our knowledge, the combined approach of methods used in this paper has not been previously applied to hyperspectral imaging analysis for interpretation and classification in the literature. The data analysis performed here aims to distinguish between four different types of plastics, some of them containing brominated flame retardants, from their near infrared hyperspectral images. The results showed that the method approach used here can be successfully used for unsupervised classification. A comparison of validation approaches, especially leave-one-out cross-validation and regions of interest scheme validation is also evaluated. Classification in hyperspectral images by independent component analysis, segmented cross-validation and uncertainty estimates

J. Spectral Imaging 7, a4 (2018). doi:10.1255/jsi.2018.a4

Independent component analysis combined with various strategies for cross-validation, uncertainty estimates by jack-knifing and critical Hotelling’s T2 limits estimation, proposed in this paper, is used for classification purposes in hyperspectral images. To the best of our knowledge, the combined approach of methods used in this paper has not been previously applied to hyperspectral imaging analysis for interpretation and classification in the literature. The data analysis performed here aims to distinguish between four different types of plastics, some of them containing brominated flame retardants, from their near infrared hyperspectral images. The results showed that the method approach used here can be successfully used for unsupervised classification. A comparison of validation approaches, especially leave-one-out cross-validation and regions of interest scheme validation is also evaluated.

]]>
Classification in hyperspectral images by independent component analysis, segmented cross-validation and uncertainty estimates Beatriz Galindo-Prieto Frank Westad doi:10.1255/jsi.2018.a4 J. Spectral Imaging 7, a4 (2018) 2018-02-25 Journal of Spectral Imaging 2018-02-25 7 1 11 10.1255/jsi.2018.a4 https://doi.org/10.1255/jsi.2018.a4
Estimation of phosphorus-based flame retardant in wood by hyperspectral imaging—a new method https://doi.org/10.1255/jsi.2018.a3 It is recognised that flame retardant chemicals degrade and leach out of flame-protected wood claddings when exposed to natural weathering. However, the ability to survey the current state of a flame retardant treatment applied to a wood cladding, an arbitrary length of time after the initial application, is limited today. In this study, hyperspectral imaging in the near infrared to short-wavelength infrared region is used to quantify the amount of flame retardant present on wooden surfaces. Several sets of samples were treated with various concentrations of a flame retardant chemical and scanned with a push broom hyperspectral camera. An inductively coupled plasma (ICP) spectroscopy analysis of the outermost layer of the treated samples was then carried out in order to determine each sample’s phosphorus content, the active ingredient in the flame retardant. Spectra from the hyperspectral images were pre-processed with extended multiplicative scatter correction, and the phosphorus content was modelled using a partial least squares (PLS) regression model. The PLS regression yielded robust predictions of surface phosphorus content with a coefficient of determination, R2, between 0.8 and 0.9 on validation data regardless of whether the flame retardant chemical had been applied to the surface of the wood or pressure-impregnated into it. The result from the study indicates that spectral imaging around the 2400–2531 nm wavelength region is favourable for quantifying the amount of phosphorus-based flame retardant contained in the outermost layer of non-coated wooden claddings. The results also reveal that the uptake of phosphorus-based flame retardant does not occur uniformly throughout the wood surface, but is to a larger extent concentrated in the earlywood regions than in the latewood. Estimation of phosphorus-based flame retardant in wood by hyperspectral imaging—a new method

J. Spectral Imaging 7, a3 (2018). doi:10.1255/jsi.2018.a3

It is recognised that flame retardant chemicals degrade and leach out of flame-protected wood claddings when exposed to natural weathering. However, the ability to survey the current state of a flame retardant treatment applied to a wood cladding, an arbitrary length of time after the initial application, is limited today. In this study, hyperspectral imaging in the near infrared to short-wavelength infrared region is used to quantify the amount of flame retardant present on wooden surfaces. Several sets of samples were treated with various concentrations of a flame retardant chemical and scanned with a push broom hyperspectral camera. An inductively coupled plasma (ICP) spectroscopy analysis of the outermost layer of the treated samples was then carried out in order to determine each sample’s phosphorus content, the active ingredient in the flame retardant. Spectra from the hyperspectral images were pre-processed with extended multiplicative scatter correction, and the phosphorus content was modelled using a partial least squares (PLS) regression model. The PLS regression yielded robust predictions of surface phosphorus content with a coefficient of determination, R2, between 0.8 and 0.9 on validation data regardless of whether the flame retardant chemical had been applied to the surface of the wood or pressure-impregnated into it. The result from the study indicates that spectral imaging around the 2400–2531 nm wavelength region is favourable for quantifying the amount of phosphorus-based flame retardant contained in the outermost layer of non-coated wooden claddings. The results also reveal that the uptake of phosphorus-based flame retardant does not occur uniformly throughout the wood surface, but is to a larger extent concentrated in the earlywood regions than in the latewood.

]]>
Estimation of phosphorus-based flame retardant in wood by hyperspectral imaging—a new method Petter Stefansson Ingunn Burud Thomas Thiis Lone Ross Gobakken Erik Larnøy doi:10.1255/jsi.2018.a3 J. Spectral Imaging 7, a3 (2018) 2018-02-15 Journal of Spectral Imaging 2018-02-15 7 1 9 10.1255/jsi.2018.a3 https://doi.org/10.1255/jsi.2018.a3
Near infrared hyperspectral imaging of blends of conventional and waxy hard wheats https://doi.org/10.1255/jsi.2018.a2 Recent development of hard winter waxy (amylose-free) wheat adapted to the North American climate has prompted the quest to find a rapid method that will determine mixture levels of conventional wheat in lots of identity preserved waxy wheat. Previous work documented the use of conventional near infrared (NIR) reflectance spectroscopy to determine the mixture level of conventional wheat in waxy wheat, with an examined range, through binary sample mixture preparation, of 0–100% (weight conventional / weight total). The current study examines the ability of NIR hyperspectral imaging of intact kernels to determine mixture levels. Twenty-nine mixtures (0, 1, 2, 3, 4, 5, 10, 15, …, 95, 96, 97, 98, 99, 100%) were formed from known genotypes of waxy and conventional wheat. Two-class partial least squares discriminant analysis (PLSDA) and statistical pattern recognition classifier models were developed for identifying each kernel in the images as conventional or waxy. Along with these approaches, conventional PLS1 regression modelling was performed on means of kernel spectra within each mixture test sample. Results indicated close agreement between all three approaches, with standard errors of prediction for the better preprocess transformations (PLSDA models) or better classifiers (pattern recognition models) of approximately 9 percentage units. Although such error rates were slightly greater than ones previously published using non-imaging NIR analysis of bulk whole kernel wheat and wheat meal, the HSI technique offers an advantage of its potential use in sorting operations. Near infrared hyperspectral imaging of blends of conventional and waxy hard wheats

J. Spectral Imaging 7, a2 (2018). doi:10.1255/jsi.2018.a2

Recent development of hard winter waxy (amylose-free) wheat adapted to the North American climate has prompted the quest to find a rapid method that will determine mixture levels of conventional wheat in lots of identity preserved waxy wheat. Previous work documented the use of conventional near infrared (NIR) reflectance spectroscopy to determine the mixture level of conventional wheat in waxy wheat, with an examined range, through binary sample mixture preparation, of 0–100% (weight conventional / weight total). The current study examines the ability of NIR hyperspectral imaging of intact kernels to determine mixture levels. Twenty-nine mixtures (0, 1, 2, 3, 4, 5, 10, 15, …, 95, 96, 97, 98, 99, 100%) were formed from known genotypes of waxy and conventional wheat. Two-class partial least squares discriminant analysis (PLSDA) and statistical pattern recognition classifier models were developed for identifying each kernel in the images as conventional or waxy. Along with these approaches, conventional PLS1 regression modelling was performed on means of kernel spectra within each mixture test sample. Results indicated close agreement between all three approaches, with standard errors of prediction for the better preprocess transformations (PLSDA models) or better classifiers (pattern recognition models) of approximately 9 percentage units. Although such error rates were slightly greater than ones previously published using non-imaging NIR analysis of bulk whole kernel wheat and wheat meal, the HSI technique offers an advantage of its potential use in sorting operations.

]]>
Near infrared hyperspectral imaging of blends of conventional and waxy hard wheats Stephen R. Delwiche Jianwei Qin Robert A. Graybosch Steven R. Rausch Moon S. Kim doi:10.1255/jsi.2018.a2 J. Spectral Imaging 7, a2 (2018) 2018-02-09 Journal of Spectral Imaging 2018-02-09 7 1 13 10.1255/jsi.2018.a2 https://doi.org/10.1255/jsi.2018.a2
When remote sensing meets topological data analysis https://doi.org/10.1255/jsi.2018.a1 Hyperspectral remote sensing plays an increasingly important role in many scientific domains and everyday life problems. Indeed, this imaging concept ends up in applications as varied as catching tax-evaders red-handed by locating new construction and building alterations, searching for aircraft and saving lives after fatal crashes, detecting oil spills for marine life and environmental preservation, spying on enemies with reconnaissance satellites, watching algae grow as an indicator of environmental health, forecasting weather to warn about natural disasters and much more. From an instrumental point of view, we can say that the actual spectrometers have rather good characteristics, even if we can always increase spatial resolution and spectral range. In order to extract ever more information from such experiments and develop new applications, we must, therefore, propose multivariate data analysis tools able to capture the shape of data sets and their specific features. Nevertheless, actual methods often impose a data model which implicitly defines the geometry of the data set. The aim of the paper is thus to introduce the concept of topological data analysis in the framework of remote sensing, making no assumptions about the global shape of the data set, but also allowing the capture of its local features. When remote sensing meets topological data analysis

J. Spectral Imaging 7, a1 (2018). doi:10.1255/jsi.2018.a1

Hyperspectral remote sensing plays an increasingly important role in many scientific domains and everyday life problems. Indeed, this imaging concept ends up in applications as varied as catching tax-evaders red-handed by locating new construction and building alterations, searching for aircraft and saving lives after fatal crashes, detecting oil spills for marine life and environmental preservation, spying on enemies with reconnaissance satellites, watching algae grow as an indicator of environmental health, forecasting weather to warn about natural disasters and much more. From an instrumental point of view, we can say that the actual spectrometers have rather good characteristics, even if we can always increase spatial resolution and spectral range. In order to extract ever more information from such experiments and develop new applications, we must, therefore, propose multivariate data analysis tools able to capture the shape of data sets and their specific features. Nevertheless, actual methods often impose a data model which implicitly defines the geometry of the data set. The aim of the paper is thus to introduce the concept of topological data analysis in the framework of remote sensing, making no assumptions about the global shape of the data set, but also allowing the capture of its local features.

]]>
When remote sensing meets topological data analysis Ludovic Duponchel doi:10.1255/jsi.2018.a1 J. Spectral Imaging 7, a1 (2018) 2018-02-06 Journal of Spectral Imaging 2018-02-06 7 1 10 10.1255/jsi.2018.a1 https://doi.org/10.1255/jsi.2018.a1