1,720,991 research outputs found
Metodi Chemiometrici per l’Analisi di Matrici Alimentari mediante Imaging Iperspettrale - Strategie per l'efficiente standardizzazione e gestione delle immagini
No full textNell’ambito dell’industria alimentare, la sempre crescente competitività di mercato determina la necessità di un continuo miglioramento dei sistemi di monitoraggio dei processi produttivi, prediligendo l’utilizzo di metodi rapidi, affidabili e non distruttivi. L’Imaging Iperspettrale (HSI) rappresenta una tecnica emergente che deriva dalla combinazione delle convenzionali tecniche di analisi dell’immagine e spettroscopiche. Questa tecnica, infatti, si basa sull’acquisizione di immagini dove ogni pixel è descritto da uno spettro completo, consentendo così di visualizzare la composizione chimica di campioni eterogenei in ogni loro punto, senza richiedere l'uso di prodotti chimici e mantenendo al tempo stesso i vantaggi di essere veloce e non distruttiva. Considerando che ogni immagine risulta quindi formata da milioni di spettri, l’analisi dei dati rappresenta la fase cruciale per l’estrazione dell’informazione utile dalla notevole quantità di dati a disposizione.
La presente tesi di dottorato ha avuto quale tema centrale lo studio di applicazioni di tecniche HSI nell’industria alimentare mediante l’utilizzo e lo sviluppo di appropriati algoritmi chemiometrici. In questo contesto, particolare attenzione è stata rivolta a due punti decisivi nell’analisi delle immagini iperspettrali: la standardizzazione e la riduzione delle dimensioni dei dati. Per quanto concerne il primo aspetto, è stata sviluppata una procedura di calibrazione interna basata sull'utilizzo di cinque materiali di riferimento che consente di minimizzare la variabilità delle immagini nel tempo e di migliorarne quindi la riproducibilità. Il tema della gestione dell’elevata quantità di dati acquisiti dai sistemi HSI è stato invece affrontato in maniera più approfondita, in quanto attualmente rappresenta uno dei maggiori limiti di queste tecniche. Il fatto che singole immagini iperspettrali di dimensione superiore a 50 MB possano essere ottenute in pochi secondi pone infatti seri problemi, sia a livello di possibili applicazioni on-line che nell’ambito della gestione dei dati durante l’elaborazione di modelli di classificazione/calibrazione. Una prima strategia è stata quindi sviluppata per ridurre le dimensioni di un'immagine training composta da milioni di spettri in una matrice ridotta formata da un ristretto numero di spettri medi, calcolati usando pixel selezionati casualmente in modo da tener conto della variabilità spaziale all'interno dell'immagine analizzata. Una strategia di riduzione dei dati più elaborata è stata invece sviluppata per la gestione e l’analisi di dataset composti da un elevato numero di immagini iperspettrali. Tale approccio consiste nel codificare le informazioni utili contenute in ciascuna immagine tridimensione in un segnale monodimensionale, denominato iperspettrogramma, che può essere visto come un'impronta digitale dell’immagine stessa. Essenzialmente, l’iperspettrogramma è ottenuto da quantità estratte mediante Analisi delle Componenti Principali ed è composto da due parti, una che codifica l'informazione spaziale e una che codifica l'informazione spettrale. Rappresentando ogni immagine con un vettore di poche centinaia di punti, tale approccio consente di analizzare con comuni metodi di analisi multivariata fino a centinaia di immagini contemporaneamente. Inoltre, mediante l’uso di algoritmi di selezione di variabili è possibile identificare le zone del segnale più importanti per la risoluzione del problema in esame, consentendo così sia l’ottimizzazione dei modelli di classificazione/calibrazione, che una efficace valutazione critica dei risultati. Infatti, le caratteristiche selezionate automaticamente dall'algoritmo possono essere sia visualizzate sotto forma di immagini che analizzate nel dominio spettrale originale. L'efficacia di questo approccio è stata dimostrata considerando vari tipi di problemi riguardanti diverse tipologie di matrici alimentari.The increasingly normative severity and market competitivity have led food industry to constantly look for improvements of process monitoring systems. In the context of fast, non-destructive and reliable techniques, image analysis-based methods have gained particular interest, thanks to their ability to spatially characterize heterogeneous samples. In particular, HyperSpectral Imaging (HSI) represents an emerging technique that provides both spatial information of imaging systems and spectral information of spectroscopy. Compared with spectroscopic methods, HSI allows to acquire the spectral data not only from a single point but at each pixel of an image, enabling the visualization of the chemical composition of the sample surface without requiring the use of chemicals, and maintaining at the same time the advantages of being fast and non-destructive. Considering that each hyperspectral image is a three-dimensional array (known as hypercube) formed by millions of spectra, the use of chemometric tools is essential to segregate only the problem-pertinent information from the huge amount of data available.
This thesis deals with the study of food industry-related applications of HSI systems, through the application and development of appropriate chemometric algorithms for the extraction of useful information from hyperspectral images. In this context, particular attention has been paid to two crucial points of HSI analysis, i.e., image standardization and data size reduction. With regard to the former issue, a novel transfer calibration procedure based on the use of five reference materials was developed, in order to minimize the variability among images over time (mainly due to instrumental instability), which can heavily affect the quality of the hyperspectral images and, therefore, the subsequent analysis results. Concerning data reduction, this issue has been more extensively faced, since the large amount of data collected by HSI systems in few seconds (individual hyperspectral images of size greater than 50 MB are usually acquired), poses serious limits to real time applications, as well as to data handling during analysis and models developments. In this context, a first compression strategy was developed to reduce a training image composed by millions of spectra into a reduced matrix formed by a restricted number of average spectra, which are calculated using randomly selected pixels in a way to account for the spatial variability within the analyzed image. A more articulated data reduction strategy has been also developed, in order to give the possibility to deal with datasets composed by a large number of hyperspectral images. This procedure consists in compressing the useful information contained in each hypercube into a one-dimensional signal, named hyperspectrogram, which is derived from quantities extracted by means of Principal Component Analysis. Essentially, the hyperspectrogram can be viewed as a fingerprint containing the relevant information brought by the hypercube, and is composed by a first part accounting for the spatial information and by a second part accounting for the spectral information. By representing each image with a vector of few hundreds of points, this procedure enables to simultaneously analyze up to hundreds of images by means of common multivariate analysis methods. Furthermore, the application to the hyperspectrograms of variable selection algorithms allows to improve the calibration/classification performances, identifying at the same time the features which are most relevant in the problem at hand. The selected features can then be evaluated both as images and in the spectral domain, allowing the interpretation in chemical terms of the features automatically selected by the algorithm. The effectiveness of this approach has been shown in the solution of different kinds of problems concerning different types of food samples
Handling large datasets of hyperspectral images: Reducing data size without loss of useful information
Get PDFHyperSpectral Imaging (HSI) is gaining increasing interest in the field of analytical chemistry, since this fast and non-destructive technique allows one to easily acquire a large amount of spectral and spatial information on a wide number of samples in very short times. However, the large size of hyperspectral image data often limits the possible uses of this technique, due to the difficulty of evaluating many samples altogether, for example when one needs to consider a representative number of samples for the implementation of on-line applications. In order to solve this problem, we propose a novel chemometric strategy aimed to significantly reduce the dataset size, which allows to analyse in a completely automated way from tens up to hundreds of hyperspectral images altogether, without losing neither spectral nor spatial information. The approach essentially consists in compressing each hyperspectral image into a signal, named hyperspectrogram, which is created by combining several quantities obtained by applying PCA to each single hyperspectral image. Hyperspectrograms can then be used as a compact set of descriptors and subjected to blind analysis techniques. Moreover, a further improvement of both data compression and calibration/classification performances can be achieved by applying proper variable selection methods to the hyperspectrograms. A visual evaluation of the correctness of the choices made by the algorithm can be obtained by representing the selected features back into the original image domain. Likewise, the interpretation of the chemical information underlying the selected regions of the hyperspectrograms related to the loadings is enabled by projecting them in the original spectral domain. Examples of applications of the hyperspectrogram-based approach to hyperspectral images of food samples in the NIR range (1000-1700 nm) and in the Vis-NIR range (400-1000 nm), facing a calibration and a defect detection issue respectively, demonstrate the effectiveness of the proposed approach
Exploration of datasets of hyperspectral images
No full textHyperspectral images of size usually greater than 50 MB can be easily acquired in very short times,
generally without the need of sample pretreatment. While Multivariate Image Analysis (MIA) tools
can be efficiently used for the exploration of single hyperspectral images or of groups composed by
a limited number (say up to 10) of merged images, the exploration of datasets composed by a large
number (>10) of images is less straightforward. However, a representative sampling of a large
number of specimens is frequently required to correctly estimate both intra- and inter-sample
variability. This implies the acquisition of datasets composed by a large number of hyperspectral
images and of several GB in size, especially in those cases where only one or a few samples can be
included in a single image scene. In this context, the exploration of the dataset by applying MIA to
single images or to subgroups of merged images does not allow to gain a global overview of the
entire dataset variability and to properly highlight the possible presence of outliers, clusters and/or
trends. A fast procedure which can be adopted to deal with this issue consists in computing the
average spectrum of each image, to build a matrix of average spectra of the analyzed hyperspectral
images. Although this approach leads sometimes to satisfactory results (especially when dealing
with homogeneous materials), the information related to spatial variability is lost, and the
hyperspectral image data are turned into “common” (i.e., not spatially resolved) spectral data. By
averaging spectra, for example, the useful information related to the presence of a defect localized
in a relatively narrow image area could be diluted within the massive amount of other “well
behaving” pixels, becoming no longer detectable.
Aiming to develop a fast and easy-to-use tool able to facilitate the exploration of large datasets of
hyperspectral images while maintaining both spectral- and spatial-related information of the
original images, we have proposed an approach which consists in automatically converting each
hyperspectral image into a signal named hyperspectrogram [1]. Essentially, the hyperspectrogram
can be viewed as a fingerprint containing the relevant information brought by the original
hyperspectral image, and is composed by a first part accounting for the spatial information and by a
second part accounting for the spectral information. By representing each image with a vector of
few hundreds of points, this procedure enables to compare simultaneously up to hundreds of images
by means of common multivariate analysis methods, such as PCA.
In order to facilitate the exploration of datasets of hyperspectral images through hyperspectrograms,
we have recently developed a Matlab Graphical User Interface (GUI), which easily allows
calculation and visualization of hyperspectrograms, exploration of the dataset and visualization of
the features of interest contained within each single sample directly in the original image domain
‘Hope for a Celestial City - A Triptych’: A musical composition for sustainability and cleaner productions for the Jing-Jin-Ji region, China
No full textThe construction of better alternatives to the present lifestyles and the solution of the environmental problems generated by modern economies and intensive exploitation of natural resources certainly require better science and technology. However, it is very unlikely that the needed global and complex solutions may be provided by one disciplinary field only, although very important, as science is. Solutions require broad minded approaches and comprehensive understanding of the multi-dimensionality of life on earth. A variety of tools and ways to approach the transition towards a cleaner and healthier environment is needed, not only to capture the complexity of life expressions, but also to value and appreciate the complexity of languages, narratives and points of view (scientific, artistic, economic, social, among others) that may lead to the “Celestial City”, a metaphor for better ways of living. This work explores the meaning and the potential contribution of a musical piece, representative of art and social languages, to inspire a transition process, with specific focus on Jing-Jin-Ji (JJJ) region, China. The piece is the ‘Hope for a Celestial City – A Triptych’ is a three-sections work for solo violin, composed by Carlotta Ferrari. It is inspired by the designed future ‘urban’ agglomerations such as the urbanised region of Jing-Jin-Ji, China. The piece refers to Bunyan's ‘Celestial City’ (this expression refers to ‘The Pilgrim's Progress’, written by John Bunyan in 1678) as a way to reflect on the past, present and future of JJJ. The work is inspired by three sets of quotes. Each of them is divided in two sub-sections: the first one is referred to as ‘literature quote’, which serves also as non-scientific reference for each section of the musical composition; the second one, on the other side, contains a set of scientific and technical references to guide and base the inspiration of the music. The first part of the piece represents the arrogant industrial civilization, “humans subdue nature”, with a particular attention to JJJ. The second part is focused on puzzling and soul-searching energy and work. In particular, the concept is that human activities can be either harmful or beneficial both for the environment and for elevating the society. The third part represents the ‘pilgrimage’ of human life and activities towards a different city: the Celestial City, where efforts achieve the result of harmonious and sustainable life. In other words, a more sustainable city can be built by transforming the present urban systems from competition to cooperation within and beyond Jing-Jin-Ji region borders. Implementing control targets, optimizing industrial structures and infrastructures, and achieving both coal consumption and air pollution reduction are seen as important technical steps towards more equitable and fulfilling life, where new patterns of connection, interaction, and exchange among residents are achieved within a healthier urban environment. The use of aesthetic inquiry, multiple languages and new creative approaches represents a support for accelerating the transition toward a more sustainable and equitable post fossil-carbon society, in particular in the case of expanding urban agglomerations. This work represents a methodological approach to artistic engagement trough music creation and performance. The purpose of such a process is to generate a collective reflection and to produce a transformative power. The collective reflection is stimulated by music performance. Furthermore, this is a way to include the contribution of artists into an important sustainability issue. Finally, this work demonstrates that music can support the communication of scientific messages, which could be difficult to be delivered in their original science form
Minimisation of instrumental noise in the acquisition of FT-NIR spectra by means of Doehlert design and signal processing techniques
No full textData reduction di immagini iperspettrali: applicazione a problemi di classificazione
No full textL'imaging iperspettrale (HSI) consente di acquisire in pochi secondi ipercubi di grandi
dimensioni, composti da milioni di spettri, che corrispondono a file spesso più grandi di 50 MB.
Questa ricchezza di dati rappresenta il principale vantaggio dell’HSI, sebbene causi seri problemi
per la gestione dei dati, tali da complicare lo sviluppo di applicazioni industriali efficienti per il
controllo in linea. Il nostro gruppo di ricerca ha recentemente proposto un’alternativa1 per trattare
dataset composti da decine o centinaia di immagini iperspettrali, che consiste nel convertire ogni
immagine iperspettrale in un segnale, chiamato iperspettrogramma, costruito in modo da
considerare sia l’informazione di natura spaziale che spettrale. Risulta così possibile trasformare
dataset composti da un elevato numero di immagini iperspettrali in matrici bidimensionali di
iperspettrogrammi, che a loro volta possono essere analizzate utilizzando i più comuni metodi
chemiometrici quali PCA, PLS o PLS-DA.
In questo contesto, presentiamo due applicazioni degli iperspettrogrammi per la soluzione di
problemi di classificazione. Una prima applicazione riguarda l'individuazione precoce di difetti
superficiali in diverse varietà di mele, con particolare attenzione ai campioni in cui il difetto non
risulta apprezzabile ad occhio nudo. Le 800 immagini iperspettrali acquisite sono state convertite in
iperspettrogrammi permettendo così la riduzione delle dimensioni del dataset da 18.6 GB a 4.7 MB.
Inoltre la selezione di variabili mediante iPLS-DA ha permesso di ridurre ulteriormente le
dimensioni del dataset e identificare le regioni più rilevanti nel segnale. Il migliore modello iPLSDA,
calcolato utilizzando solo 30 variabili delle 1200 iniziali, ha portato ad un valore di efficienza
in predizione sul test set esterno pari a 89.6%. Una seconda applicazione riguarda la classificazione
di caffè verde appartenente a diverse tipologie: Arabica e Robusta. Prove preliminari hanno
mostrato come la classificazione mediante PLS-DA effettuata sugli iperspettrogrammi ha portato ad
un valore di efficienza in predizione del 98.3%
Fast exploration and classification of large hyperspectral image datasets for early bruise detection on apples
No full textHyperspectral imaging allows to easily acquire tens of thousands of spectra for a single sample in few seconds; though valuable, this data-richness poses many problems due to the difficulty of handling a representative amount of samples altogether. For this reason, we recently proposed an approach based on the idea of reducing each image into a one-dimensional signal, named hyperspectrogram, which accounts both for spatial and for spectral information. In this manner, a dataset of hyperspectral images can be easily and quickly converted into a set of signals (2D data matrix), which in turn can be analyzed using classical chemometric techniques. In this work, the hyperspectrograms obtained from a dataset of 800 NIR-hyperspectral images of two different apple varieties were used to discriminate bruised from sound apples using iPLS-DA as variable selection algorithm, which allowed to efficiently detect the presence of bruises. Moreover, the reconstruction as images of the selected variables confirmed that the automated procedure led to the exact identification of the spatial features related to the onset and to the subsequent evolution with time of the bruise defect
- …
