En fournissant en plus de l'information spatiale une mesure spectrale en fonction des longueurs d'ondes, l'imagerie hyperspectrale s'enorgueillie d'atteindre une précision bien plus importante que l'imagerie couleur. Grâce à cela, elle a été utilisée en contrôle qualité, inspection de matériaux,… Cependant, pour exploiter pleinement ce potentiel, il est important de traiter la donnée spectrale comme une mesure, d'où la nécessité de la métrologie, pour laquelle exactitude, incertitude et biais doivent être maitrisés à tous les niveaux de traitement.Face à cet objectif, nous avons choisi de développer une approche non-linéaire, basée sur la morphologie mathématique et de l'étendre au domaine spectral par le biais d'une relation d'ordre spectral basée sur les fonctions de distance. Une nouvelle fonction de distance spectrale et une nouvelle relation d'ordonnancement sont ainsi proposées. De plus, un nouvel outil d'analyse du basé sur les histogrammes de différences spectrales a été développé.Afin d'assurer la validité des opérateurs, une validation théorique rigoureuse et une évaluation métrologique ont été mises en œuvre à chaque étage de développement. Des protocoles d'évaluation de la qualité des traitements morphologiques sont proposés, exploitant des jeux de données artificielles pour la validation théorique, des ensembles de données dont certaines caractéristiques sont connues pour évaluer la robustesse et la stabilité et des jeux de données de cas réel pour prouver l'intérêt des approches en contexte applicatif. Les applications sont développées dans le contexte du patrimoine culturel pour l'analyse de peintures et pigments. / Providing not only spatial information but also spectral measure as a function of wavelength, hyperspectral imaging boasts a much greater gain in accuracy than the traditional color imaging. And for this capability, hyperspectral imaging has been employed for quality control, inspection of materials in various fields. However, to fully exploit this potential, it is important to process the spectral data as a measure. This induces the need of metrology where accuracy, uncertainty, and bias are managed at every level of processing.Aiming at developing a metrological image processing framework for spectral data, we select to develop a nonlinear approach using the mathematical morphology framework and extended it to the spectral domain by means of a distance-based ordering relation. A novel spectral distance function and spectral ordering relation are proposed, in addition of a new analysis tools based on spectral differences. To ensure the validity of the spectral mathematical morphology framework, rigorous theoretical validation and metrological assessment are carried out at each development stages. So, protocols for quality assessment of spectral image processing tools are developed. These protocols consist of artificial datasets to validate completely the theoretical requirements, datasets with known characteristics to assess the robustness and stability, and datasets from real cases to proof the usefulness of the framework on applicative context. The application tasks themselves are within the cultural heritage domain, where the target images come from pigments and paintings. / Hyperspektral avbildning muliggjør mye mer nøyaktige målinger enn tradisjonelle gråskala og fargebilder, gjennom både høy romlig og spektral oppløsning (funksjon av bølgelengde). På grunn av dette har hyperspektral avbildning blitt anvendt i økende grad ulike applikasjoner som kvalitetskontroll og inspeksjon av materialer. Men for å fullt ut utnytte sitt potensiale, er det viktig å være i stand til å behandle spektrale bildedata som målinger på en gyldig måte. Dette induserer behovet for metrologi, der nøyaktighet, usikkerhet og skjevhet blir adressert og kontrollert på alle nivå av bildebehandlingen.Med sikte på å utvikle et metrologisk rammeverk for spektral bildebehandling valgte vi en ikke-lineær metodikk basert på det etablerte matematisk morfologi-rammeverket. Vi har utvidet dette rammeverket til det spektrale domenet ved hjelp av en avstandsbasert sorteringsrelasjon. En ny spektral avstandsfunksjon og nye spektrale sorteringsrelasjoner ble foreslått, samt nye verktøy for spektral bildeanalyse basert på histogrammer av spektrale forskjeller.For å sikre gyldigheten av det nye spektrale rammeverket for matematisk morfologi, har vi utført en grundig teoretisk validering og metrologisk vurde-ring på hvert trinn i utviklingen. Dermed er og-så nye protokoller for kvalitetsvurdering av spektrale bildebehandlingsverktøy utviklet. Disse protokollene består av kunstige datasett for å validere de teoretiske måletekniske kravene, bildedatasett med kjente egenskaper for å vurdere robustheten og stabiliteten, og datasett fra reelle anvendelser for å bevise nytten av rammeverket i en anvendt sammenheng. De valgte anvendelsene er innenfor kulturminnefeltet, hvor de analyserte bildene er av pigmenter og malerier.

A plasticidade do Eucalyptus às mudanças no clima merece destaque, principalmente pelo fato da maioria dos povoamentos florestais implantados no Brasil serem, em geral, estabelecidos em regiões de baixa fertilidade, pobres em potássio (K) e submetidos a longos períodos de déficit hídrico. A expansão das plantações pode ser prejudicada pelas mudanças climáticas, com a alteração da frequência e intensidade da precipitação. Desta forma, o presente trabalho tem como objetivo avaliar o efeito da interação do déficit hídrico e da fertilização com K e sódio (Na) no crescimento e qualidade do lenho de árvores de E. grandis sob o contexto das mudanças climáticas. Em experimento do tipo split-plot instalado em junho de 2010, foram selecionadas 54 árvores de eucalipto com cinco anos submetidas a dois regimes hídricos (100% e 63%) e três fertilizantes (K, Na e controle). Realizou-se a amostragem do lenho, coletando-se seis discos nas posições base, DAP, 25, 50, 75 e 100% da altura total. Avaliou-se as propriedades físicas (densidade aparente por densitometria de raios x e predição da densidade básica por NIR), propriedades anatômicas (fibras e vasos) e química (predição do teor de extrativos totais pela aquisição da imagem HI-NIR), de acordo com os tratamentos. Para a predição dos extrativos, propôs-se a predição direta e a transferência de calibração entre equipamentos NIR e HI-NIR. A transferência foi baseada em uma coleção de base de calibração completa medida nos dois aparelhos. Comparou-se quatro modelos de transferência de calibração (Update, Repfile, PDS e TOP). A eficácia dos modelos foi testada em um grupo de amostras teste (1/3 das amostras totais). Os resultados mostram que em todas as propriedades do lenho houve efeito significativo dos tratamentos; a densidade aparente é menor nas árvores fertilizadas com K e Na e sem exclusão parcial de chuvas. A predição da densidade básica apresentou resultado satisfatório com RMSECV igual a 0,022 g/cm³. As fibras são maiores nas árvores fertilizadas com K e apresentam maior espessura de parede nas árvores controle, ambas na condição de exclusão parcial de chuvas. Os vasos e a largura dos anéis de crescimento sofreram influência, principalmente, da exclusão de chuvas. A predição direta do teor de extrativos totais foi eficácia, mostrando resultados semelhantes com a literatura para valores de extrativos de eucalipto preditos e observados. O tratamento K sem exclusão parcial de chuvas apresentou o menor valor médio predito de extrativos totais (3,90%). O melhor modelo de transferência de calibração foi o TOP, com SEP de 1,53%, SECV de 1,41% e R² de 0,88. Conclui-se que a interação do déficit hídrico e da fertilização influenciou as propriedades do lenho das árvores de E. grandis e as análises realizadas permitem traçar estratégias mais adequadas para dar subsídio à expansão de povoamentos florestais brasileiros em áreas sujeitas a longos períodos de seca. / The plasticity of the Eucalyptus to changes in the climate deserves to be highlighted, mainly because most of the forests implanted in Brazil are generally established in regions of low fertility, poor in potassium (K) and subjected to long periods of drought. Expansion of plantations can be hindered by climate change, with changes in the frequency and intensity of precipitation. The objective of this work is to evaluate the interaction of drought and fertilization with K and sodium (Na) on the growth and quality of E. grandis trees in the context of climate change. In a split-plot experiment installed in June 2010, 54 Eucalyptus trees with five years submitted to two water regimes (100% and 63%) and three fertilizers (K, Na and control) were selected. Sampling was carried out by collecting six discs at the base positions, DBH, 25, 50, 75 and 100% of the total height. The physical properties (apparent density by x-ray densitometry and basic density prediction by NIR), anatomical properties (fibers and vessels) and chemistry (prediction of total extractive content by HI-NIR image acquisition) were evaluated according to the treatments. For the prediction of extractives, the direct prediction and calibration transfer between NIR and HI-NIR equipment was proposed. The transfer was based on a complete calibration base collection measured on both devices. Four calibration transfer models (Update, Repfile, PDS and TOP) were compared. The efficacy of the models was tested in test set samples (1/3 of the total samples). The results show that in all the properties of the wood there was significant effect of the treatments; the apparent density is lower in the trees fertilized with K and Na and without partial throughfall exclusion. The prediction of the basic density presented satisfactory results with RMSECV of 0.022 g/cm³. The fibers are larger in the trees fertilized with K and present a greater thickness of wall in the control trees, both in partial throughfall exclusion. The vessels and the width of the growth rings were influenced, mainly, by partial throughfall exclusion. The direct prediction of the total extractive content was efficacy, showing similar results with the literature for values of predicted and observed Eucalyptus extractives. The K treatment without partial rainfall exclusion had the lowest predicted mean value of total extractives (3.90%). The best calibration transfer model was the TOP, with SEP of 1.53%, SECV of 1.41% and R² of 0.88. The conclusion of this work is that the interaction of water deficit and fertilization influences the wood properties of E. grandis trees and the analyzes carried out allow to draw up more adequate strategies to subsidize the expansion of Brazilian plantation forests in areas subject to long periods of drought.

A plasticidade do Eucalyptus às mudanças no clima merece destaque, principalmente pelo fato da maioria dos povoamentos florestais implantados no Brasil serem, em geral, estabelecidos em regiões de baixa fertilidade, pobres em potássio (K) e submetidos a longos períodos de déficit hídrico. A expansão das plantações pode ser prejudicada pelas mudanças climáticas, com a alteração da frequência e intensidade da precipitação. Desta forma, o presente trabalho tem como objetivo avaliar o efeito da interação do déficit hídrico e da fertilização com K e sódio (Na) no crescimento e qualidade do lenho de árvores de E. grandis sob o contexto das mudanças climáticas. Em experimento do tipo split-plot instalado em junho de 2010, foram selecionadas 54 árvores de eucalipto com cinco anos submetidas a dois regimes hídricos (100% e 63%) e três fertilizantes (K, Na e controle). Realizou-se a amostragem do lenho, coletando-se seis discos nas posições base, DAP, 25, 50, 75 e 100% da altura total. Avaliou-se as propriedades físicas (densidade aparente por densitometria de raios x e predição da densidade básica por NIR), propriedades anatômicas (fibras e vasos) e química (predição do teor de extrativos totais pela aquisição da imagem HI-NIR), de acordo com os tratamentos. Para a predição dos extrativos, propôs-se a predição direta e a transferência de calibração entre equipamentos NIR e HI-NIR. A transferência foi baseada em uma coleção de base de calibração completa medida nos dois aparelhos. Comparou-se quatro modelos de transferência de calibração (Update, Repfile, PDS e TOP). A eficácia dos modelos foi testada em um grupo de amostras teste (1/3 das amostras totais). Os resultados mostram que em todas as propriedades do lenho houve efeito significativo dos tratamentos; a densidade aparente é menor nas árvores fertilizadas com K e Na e sem exclusão parcial de chuvas. A predição da densidade básica apresentou resultado satisfatório com RMSECV igual a 0,022 g/cm³. As fibras são maiores nas árvores fertilizadas com K e apresentam maior espessura de parede nas árvores controle, ambas na condição de exclusão parcial de chuvas. Os vasos e a largura dos anéis de crescimento sofreram influência, principalmente, da exclusão de chuvas. A predição direta do teor de extrativos totais foi eficácia, mostrando resultados semelhantes com a literatura para valores de extrativos de eucalipto preditos e observados. O tratamento K sem exclusão parcial de chuvas apresentou o menor valor médio predito de extrativos totais (3,90%). O melhor modelo de transferência de calibração foi o TOP, com SEP de 1,53%, SECV de 1,41% e R² de 0,88. Conclui-se que a interação do déficit hídrico e da fertilização influenciou as propriedades do lenho das árvores de E. grandis e as análises realizadas permitem traçar estratégias mais adequadas para dar subsídio à expansão de povoamentos florestais brasileiros em áreas sujeitas a longos períodos de seca. / The plasticity of the Eucalyptus to changes in the climate deserves to be highlighted, mainly because most of the forests implanted in Brazil are generally established in regions of low fertility, poor in potassium (K) and subjected to long periods of drought. Expansion of plantations can be hindered by climate change, with changes in the frequency and intensity of precipitation. The objective of this work is to evaluate the interaction of drought and fertilization with K and sodium (Na) on the growth and quality of E. grandis trees in the context of climate change. In a split-plot experiment installed in June 2010, 54 Eucalyptus trees with five years submitted to two water regimes (100% and 63%) and three fertilizers (K, Na and control) were selected. Sampling was carried out by collecting six discs at the base positions, DBH, 25, 50, 75 and 100% of the total height. The physical properties (apparent density by x-ray densitometry and basic density prediction by NIR), anatomical properties (fibers and vessels) and chemistry (prediction of total extractive content by HI-NIR image acquisition) were evaluated according to the treatments. For the prediction of extractives, the direct prediction and calibration transfer between NIR and HI-NIR equipment was proposed. The transfer was based on a complete calibration base collection measured on both devices. Four calibration transfer models (Update, Repfile, PDS and TOP) were compared. The efficacy of the models was tested in test set samples (1/3 of the total samples). The results show that in all the properties of the wood there was significant effect of the treatments; the apparent density is lower in the trees fertilized with K and Na and without partial throughfall exclusion. The prediction of the basic density presented satisfactory results with RMSECV of 0.022 g/cm³. The fibers are larger in the trees fertilized with K and present a greater thickness of wall in the control trees, both in partial throughfall exclusion. The vessels and the width of the growth rings were influenced, mainly, by partial throughfall exclusion. The direct prediction of the total extractive content was efficacy, showing similar results with the literature for values of predicted and observed Eucalyptus extractives. The K treatment without partial rainfall exclusion had the lowest predicted mean value of total extractives (3.90%). The best calibration transfer model was the TOP, with SEP of 1.53%, SECV of 1.41% and R² of 0.88. The conclusion of this work is that the interaction of water deficit and fertilization influences the wood properties of E. grandis trees and the analyzes carried out allow to draw up more adequate strategies to subsidize the expansion of Brazilian plantation forests in areas subject to long periods of drought.

Dans cette thèse, nous avons présenté les aspects de la technologie d'imagerie hyperspectrale en concentrant sur le problème de démélange non-linéaire. Pour cette tâche, nous avons proposé trois solutions. La première consiste à intégrer les avantages de l'apprentissage de variétés dans les méthodes de démélange classique pour concevoir leurs versions non-linéaires. Les résultats avec les données générées sur une variété bien connue - le "Swissroll"- donne des résultats prometteurs. Les méthodes fonctionnent beaucoup mieux avec l'augmentation de la non-linéarité. Cependant, l'absence de contrainte de non-négativité dans ces méthodes reste une question ouverte pour des améliorations à trouver. La deuxième proposition vise à utiliser la méthode de pré-image pour estimer une transformation inverse de l'espace de données entrées des pixels vers l'espace des abondances. L'ajout des informations spatiales sous forme "variation totale" est également introduit pour rendre l'algorithme plus robuste au bruit. Néanmoins, le problème d'obtention des données de réalité terrain nécessaires pour l'étape d'apprentissage limite l'application de ce type d'algorithmes. / In this thesis , we present several aspects of hyperspectral imaging technology , while focusing on the problem of non- linear unmixing . We have proposed three solutions for this task. The first one is integrating the advantages of manifold learning in classical unmixing methods to design their nonlinear versions . Results with data generated on a well-known manifold- the " Swissroll " - seem promising. The methods work much better with the increase in non- linearity compared with their linear version. However, the absence of constraint of non- negativity in these methods remains an open question for improvements . The second proposal is using the pre-image method for estimating an inverse transformation of the data form pixel space to abundance of space . The adoption of spatial information as " total variation " is also introduced to make the algorithm more robust to noise . However, the problem of obtaining ground truth data required for learning step limits the application of such algorithms.

The airborne hyperspectral remote sensing is used as an approach to monitor and analyse actual state of environmental components. This thesis deals with hyperspectral image data, especially it is focused on normalization with respect to scanning angle. The thesis proposes specific algorithm, which is based on the statistical analysis of spectral lines across the scan line and on a physical models describing the process of spectral reflectance. An important part of this thesis is software implementation of proposed algorithm, that allows to calculate required normalization for real datasets.

No description available.

No description available.

Development of selected invasive species and meadow vegetation classification algorithm in the Krkonoše Mountains using hyperspectral data Abstract The thesis deals with utilization of airbone APEX hyperspectral image data for selected invasive species and meadow vegetation classification in the study area of the Krkonoše Mountains National Park. The mian goal of the thesis was to develop of classification algorithm based on proposed vegetation indices. The approach was based on the utilization of in-situ LAI, fAPAR, chlorophyll content data and analysis of their relation with vegetation spectral properties. The work also deals with several problems regarding LAI - vegetation indices relationship, namely saturation of LAI and mutual correlation of LAI and chlorophyll content. Tha classification was focued on invasive species Rumex alpinus and Lupinus polyphyllus, meadow vegetation with dominant Nardus stricta and dominant Trisetum flavescens and cutted lawns. Besides the proposed approach, the presented work resulted in several classification maps of study area and in spectral libraries, containing ground level spectra of studied invasive species, meadow vegetation types and several other meadow species. Keywords: hyperspectral image data, APEX, LAI, fAPAR, vegetation indices, invasive species, meadow...

In this dissertation, some new algorithms are developed for hyperspectral imaging analysis enhancement. Tensor data format is applied in hyperspectral dataset sparse and low-rank decomposition, which could enhance the classification and detection performance. And multi-view learning technique is applied in hyperspectral imaging clustering. Furthermore, kernel version of multi-view learning technique has been proposed, which could improve clustering performance. Most of low-rank and sparse decomposition algorithms are based on matrix data format for HSI analysis. As HSI contains high spectral dimensions, tensor based extended low-rank and sparse decomposition (TELRSD) is proposed in this dissertation for better performance of HSI classification with low-rank tensor part, and HSI detection with sparse tensor part. With this tensor based method, HSI is processed in 3D data format, and information between spectral bands and pixels maintain integrated during decomposition process. This proposed algorithm is compared with other state-of-art methods. And the experiment results show that TELRSD has the best performance among all those comparison algorithms. HSI clustering is an unsupervised task, which aims to group pixels into different groups without labeled information. Low-rank sparse subspace clustering (LRSSC) is the most popular algorithms for this clustering task. The spatial-spectral based multi-view low-rank sparse subspace clustering (SSMLC) algorithms is proposed in this dissertation, which extended LRSSC with multi-view learning technique. In this algorithm, spectral and spatial views are created to generate multi-view dataset of HSI, where spectral partition, morphological component analysis (MCA) and principle component analysis (PCA) are applied to create others views. Furthermore, kernel version of SSMLC (k-SSMLC) also has been investigated. The performance of SSMLC and k-SSMLC are compared with sparse subspace clustering (SSC), low-rank sparse subspace clustering (LRSSC), and spectral-spatial sparse subspace clustering (S4C). It has shown that SSMLC could improve the performance of LRSSC, and k-SSMLC has the best performance. The spectral clustering has been proved that it equivalent to non-negative matrix factorization (NMF) problem. In this case, NMF could be applied to the clustering problem. In order to include local and nonlinear features in data source, orthogonal NMF (ONMF), graph-regularized NMF (GNMF) and kernel NMF (k-NMF) has been proposed for better clustering performance. The non-linear orthogonal graph NMF combine both kernel, orthogonal and graph constraints in NMF (k-OGNMF), which push up the clustering performance further. In the HSI domain, kernel multi-view based orthogonal graph NMF (k-MOGNMF) is applied for subspace clustering, where k-OGNMF is extended with multi-view algorithm, and it has better performance and computation efficiency.

<p dir="ltr">We explore the efficacy of sparsity and ensemble model in the classification of hyperspectral images, a pivotal task in remote sensing applications. While Convolutional Neural Networks (CNNs) and Transformer models have shown promise in this domain, each exhibits distinct limitations; CNNs excel in capturing the spatial/local features but falter to capture spectral features, whereas Transformers captures the spectral features at the expense of spatial features. Furthermore, the computational cost associated with training several independent CNN and Transformer networks becomes expensive. To address these limitations, we propose a novel ensemble framework comprising pruned CNNs and Transformers, optimizing both spatial and spectral feature utilization while curbing computational costs. By integrating sparsity through model pruning, our approach effectively reduces redundancy and computational complexity without compromising accuracy. Through extensive experimentation, we find that our method achieves comparable accuracy to its non-sparse counterparts while decreasing the computational cost. Our contribution enhances remote sensing analytics by demonstrating the potential of sparse and ensemble models in improving the precision and computational efficiency of hyperspectral image classification.</p>