La finesse de la résolution spectrale des images hyperspectrales en télédétection permet une analyse précise de la scène observée, mais leur résolution spatiale est limitée, et un pixel acquis par le capteur est souvent un mélange des contributions de différents matériaux. Le démélange spectral permet d'estimer les spectres des matériaux purs (endmembers) de la scène, et leurs abondances dans chaque pixel. Les endmembers sont souvent supposés être parfaitement représentés par un seul spectre, une hypothèse fausse en pratique, chaque matériau ayant une variabilité intra-classe non négligeable. Le but de cette thèse est de développer des algorithmes prenant mieux en compte ce phénomène. Nous effectuons le démélange localement, dans des régions bien choisies de l'image où les effets de la variabilité sont moindres, en éliminant automatiquement les endmembers non pertinents grâce à de la parcimonie collaborative. Dans une autre approche, nous raffinons l'estimation des abondances en utilisant la structure de groupe d'un dictionnaire d'endmembers extrait depuis les données. Ensuite, nous proposons un modèle de mélange linéaire étendu, basé sur des considérations physiques, qui modélise la variabilité spectrale par des facteurs d'échelle, et développons des algorithmes d'optimisation pour en estimer les paramètres. Ce modèle donne des résultats facilement interprétables et de meilleures performances que d'autres approches de la littérature. Nous étudions enfin deux applications de ce modèle pour confirmer sa pertinence. / The fine spectral resolution of hyperspectral remote sensing images allows an accurate analysis of the imaged scene, but due to their limited spatial resolution, a pixel acquired by the sensor is often a mixture of the contributions of several materials. Spectral unmixing aims at estimating the spectra of the pure materials (called endmembers) in the scene, and their abundances in each pixel. The endmembers are usually assumed to be perfectly represented by a single spectrum, which is wrong in practice since each material exhibits a significant intra-class variability. This thesis aims at designing unmixing algorithms to better handle this phenomenon. First, we perform the unmixing locally in well chosen regions of the image where variability effects are less important, and automatically discard wrongly estimated local endmembers using collaborative sparsity. In another approach, we refine the abundance estimation of the materials by taking into account the group structure of an image-derived endmember dictionary. Second, we introduce an extended linear mixing model, based on physical considerations, modeling spectral variability in the form of scaling factors, and develop optimization algorithms to estimate its parameters. This model provides easily interpretable results and outperforms other state-of-the-art approaches. We finally investigate two applications of this model to confirm its relevance.

En imagerie hyperspectrale, chaque pixel est associé à un spectre provenant de la réflectance observée en d points de mesure (i.e., longueurs d'onde). On se retrouve souvent dans une situation où la taille d'échantillon n est relativement faible devant le nombre d de variables. Ce phénomène appelé "fléau de la dimension" est bien connu en statistique multivariée. Plus d augmente devant n, plus les performances des méthodologies statistiques standard se dégradent. Les spectres de réflectance intègrent dans leur dimension spectrale un continuum qui leur confère une nature fonctionnelle. Un hyperspectre peut être modélisé par une fonction univariée de la longueur d'onde, sa représentation produisant une courbe. L'utilisation de méthodes fonctionnelles sur de telles données permet de prendre en compte des aspects fonctionnels tels que la continuité, l'ordre des bandes spectrales, et de s'affranchir des fortes corrélations liées à la finesse de la grille de discrétisation. L'objectif principal de cette thèse est d'évaluer la pertinence de l'approche fonctionnelle dans le domaine de la télédétection hyperspectrale lors de l'analyse statistique. Nous nous sommes focalisés sur le modèle non-paramétrique de régression fonctionnelle, couvrant la classification supervisée. Dans un premier temps, l'approche fonctionnelle a été comparée avec des méthodes multivariées usuellement employées en télédétection. L'approche fonctionnelle surpasse les méthodes multivariées dans des situations délicates où l'on dispose d'une petite taille d'échantillon d'apprentissage combinée à des classes relativement homogènes (c'est-à-dire difficiles à discriminer). Dans un second temps, une alternative à l'approche fonctionnelle pour s'affranchir du fléau de la dimension a été développée à l'aide d'un modèle parcimonieux. Ce dernier permet, à travers la sélection d'un petit nombre de points de mesure, de réduire la dimensionnalité du problème tout en augmentant l'interprétabilité des résultats. Dans un troisième temps, nous nous sommes intéressés à la situation pratique quasi-systématique où l'on dispose de données fonctionnelles contaminées. Nous avons démontré que pour une taille d'échantillon fixée, plus la discrétisation est fine, meilleure sera la prédiction. Autrement dit, plus d est grand devant n, plus la méthode statistique fonctionnelle développée est performante. / In hyperspectral imaging, each pixel is associated with a spectrum derived from observed reflectance in d measurement points (i.e., wavelengths). We are often facing a situation where the sample size n is relatively low compared to the number d of variables. This phenomenon called "curse of dimensionality" is well known in multivariate statistics. The mored increases with respect to n, the more standard statistical methodologies performances are degraded. Reflectance spectra incorporate in their spectral dimension a continuum that gives them a functional nature. A hyperspectrum can be modelised by an univariate function of wavelength and his representation produces a curve. The use of functional methods allows to take into account functional aspects such as continuity, spectral bands order, and to overcome strong correlations coming from the discretization grid fineness. The main aim of this thesis is to assess the relevance of the functional approach in the field of hyperspectral remote sensing for statistical analysis. We focused on the nonparametric fonctional regression model, including supervised classification. Firstly, the functional approach has been compared with multivariate methods usually involved in remote sensing. The functional approach outperforms multivariate methods in critical situations where one has a small training sample size combined with relatively homogeneous classes (that is to say, hard to discriminate). Secondly, an alternative to the functional approach to overcome the curse of dimensionality has been proposed using parsimonious models. This latter allows, through the selection of few measurement points, to reduce problem dimensionality while increasing results interpretability. Finally, we were interested in the almost systematic situation where one has contaminated functional data. We proved that for a fixed sample size, the finer the discretization, the better the prediction. In other words, the larger dis compared to n, the more effective the functional statistical methodis.

Orientador: Maria de Lourdes Bueno Trindade Galo / Banca: Renata Ribeiro de Araújo Rocha / Banca: Waterloo Pereira Filho / Resumo: Os recentes avanços na tecnologia de Sensoriamento Remoto proporcionaram o desenvolvimento de sensores orbitais com altíssima resolução espectral, capazes de fornecer medidas radiométricas em bandas estreitas e contínuas para cada pixel da imagem, definindo curvas espectrais com potencial de discriminar diferentes componentes da matéria. Diante disso, o objetivo geral deste trabalho foi de avaliar a contribuição de imagens hiperespectrais na identificação de componentes opticamente ativos presentes em um corpo d'água, considerado de boa qualidade. Para tanto, uma imagem hiperespectral Hyperion foi adquirida simultaneamente com variáveis limnológicas coletadas em alguns pontos georreferenciados no reservatório de Itupararanga. Após a correção atmosférica da imagem, extraíram curvas espectrais nos locais geográficos dos pontos, nos quais se aplicaram técnicas de análise de espectros, tais como, remoção do contínuo, razão de bandas e análise derivativa. Os dados hiperespectrais originais e os resultantes da aplicação de técnicas foram correlacionados com algumas variáveis limnológicas... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: The recent improvements in the technology of Remote Sensing are providing the development of sensors with high spectral resolution that can supply radiometric measurements in narrow and continuous bands for each pixel of the image, defining spectral curves with potential of separating several components of the matter. Due to that, the general objective of this work was evaluating the contribution of hyperspectral images in the identification of optically active constituents present in a body of water, considered good quality. To reach the proposed objective, a hyperspectral imagery of EO-1/Hyperion orbital sensor was acquired simultaneously with limnological variables collected in some points in the body of water. After correcting the atmospheric effects, in the geographical locations of those points, spectral curves of the hyperspectral image were extracted, in which techniques of spectral analysis were applied, such as, continuum removal, derivative analysis and ratio analysis. The hyperspectral original data and the resultants of the application of techniques were correlated with some limnological variables. Of the applied techniques, the derivative analysis provided better differentiation among the optically active constituents... (Complete abstract, click electronic access below) / Mestre

La connaissance des matériaux de surface est essentielle pour l’aménagement et la gestion des villes. Avec les avancées en télédétection, particulièrement en imagerie de haute résolution spatiale et spectrale, l’identification et la cartographie détaillée des matériaux de surface en milieu urbain sont maintenant envisageables. Les signatures spectrales décrivent les interactions entre les objets au sol et le rayonnement solaire, et elles sont supposées uniques pour chaque type de matériau de surface. Dans ce projet de recherche nous avons utilisé des images hyperspectrales aériennes du capteur CASI, avec une résolution de 1 m2 et 96 bandes contigües entre 380nm et 1040nm. Ces images couvrant l’île de Montréal (QC, Canada), acquises en 2016, ont été analysées pour identifier les matériaux de surfaces. Pour atteindre ces objectifs, notre méthode d’analyse est fondée sur la comparaison des signatures spectrales d’un pixel quelconque à celles des objets typiques contenues dans des bibliothèques spectrales (matériaux inertes et végétation). Pour mesurer la correspondance entre la signature spectrale d’un objet et la signature spectrale de référence nous avons utilisé deux métriques. La première métrique tient compte de la forme d’une signature spectrale et la seconde, de la différence des valeurs de réflectance entre la signature spectrale observée et celle de référence. Un classificateur flou utilisant ces deux métriques est alors appliqué afin de reconnaître le type de matériau de surface sur la base du pixel. Des signatures spectrales typiques ont été extraites des deux librairies spectrales (ASTER et HYPERCUBE). Des signatures spectrales des objets typiques à Montréal mesurées sur le terrain (spectroradiomètre ASD) ont été aussi utilisées comme références. Trois grandes catégories de matériaux ont été identifiées dans les images pour faciliter la comparaison entre les classifications par source de références spectrales : l’asphalte, le béton et la végétation. La classification utilisant ASTER comme bibliothèque de référence a eu le plus grand taux de réussite avec 92%, suivi par ASD à 88% et finalement HYPERCUBE avec 80%. Nous 5 n’avons pas trouvé de différences significatives entre les trois résultats, ce qui indique que la classification est indépendante de la source des signatures spectrales de référence. / Knowledge of surface cover materials is crucial for urban planning and management. With advances in remote sensing, especially in high spatial and spectral resolution imagery, the identification and detailed mapping of surface materials in urban areas based on spectral signatures are now feasible. Spectral signatures describe the interactions between ground objects and solar radiation and are assumed unique for each type of material. In this research, we use airborne CASI images with 1 m2 spatial resolution, with 96 contiguous bands in a spectral range between 367 nm and 1044 nm. These images covering the island of Montreal (Quebec, Canada), obtained in 2016, were analyzed to identify urban surface materials. The objectives of the project were first to find a correspondence between the physical and chemical characteristic of typical surface materials, present in the Montreal scenes, and the spectral signatures within the images. Second, to develop a sound methodology for identifying these surface materials in urban landscapes. To reach these objectives, our method of analysis is based on a comparison of pixel spectral signatures to those contained in a reference spectral library that describe typical surface covering materials (inert materials and vegetation). Two metrics were used in order to measure the correspondence of pixel spectral signatures and reference spectral signature. The first metric considers the shape of a spectral signature and the second the difference of reflectance values between the observed and reference spectral signature. A fuzzy classifier using these two metrics is then applied to recognize the type of material on a pixel basis. Typical spectral signatures were extracted from two spectral libraries (ASTER and HYPERCUBE). Spectral signatures of typical objects in Montreal measured on the ground (ASD spectroradiometer) were also used as reference spectra. Three general types of surface materials (asphalt, concrete, and vegetation) were used to ease the comparison between classifications using these spectral libraries. The classification using ASTER as a reference library had the highest success rate reaching 92%, followed by the field spectra at 88%, and finally with HYPERCUBE at 80%. There were no significant differences in the classification results indicating that the methodology works independently of the source of reference spectral signatures.

No description available.

Conselho Nacional de Desenvolvimento CientÃfico e TecnolÃgico / A caracterizaÃÃo, delimitaÃÃo e avaliaÃÃo das Ãreas afetadas por sais/sÃdio à de extrema relevÃncia para o PerÃmetro de IrrigaÃÃo de Morada Nova â Cearà podendo contribuir nas tomadas de decisÃes referentes à exploraÃÃo agrÃcola local. O sensoriamento remoto (SR) pode ser uma alternativa atraente para complementar o uso de mÃtodos tradicionais em funÃÃo de seu baixo custo, ampla cobertura espacial, frequÃncia temporal de aquisiÃÃo de imagens, visando possibilitar o mapeamento das Ãreas salinizadas. O presente trabalho teve como objetivo usar dados de SR no desenvolvimento de estratÃgias metodolÃgicas para identificar Ãreas com problemas de salinidade, visando avaliaÃÃes dos efeitos provocados no solo e na vegetaÃÃo. Inicialmente, foi usada espectroscopia de reflectÃncia de laboratÃrio para caracterizar e quantificar variaÃÃes na reflectÃncia e nas bandas de absorÃÃo espectrais em funÃÃo das alteraÃÃes da condutividade elÃtrica (CE) do solo, um indicador indireto de salinizaÃÃo. Amostras de Neossolos FlÃvicos (n =180) foram salinizadas com crescentes nÃveis de NaCl, MgCl2 e CaCl2. Metade das amostras foi tratada com gesso agrÃcola, um corretivo de salinizaÃÃo dos solos comumente utilizado na regiÃo. Espectros de reflectÃncia foram medidos ao nadir em ambiente controlado (laboratÃrio) usando o espectrÃmetro FieldSpec. As variaÃÃes de reflectÃncia e absorÃÃo foram avaliadas atravÃs da anÃlise por componentes principais (ACP) e da tÃcnica remoÃÃo do contÃnuo (RC), respectivamente, para solos tratados com gesso (TG) e nÃo tratados com gesso (NTG). Usando parte das amostras NTG (n = 62) e um conjunto de amostras independentes (n = 32), coletadas em vÃrios pontos dentro do perÃmetro irrigado, modelos preditivos foram desenvolvidos usando regressÃes lineares de bandas individuais do espectrÃmetro, Ãndice normalizado de salinidade (NDSI) e regressÃo por mÃnimos quadrados parciais (PLSR). Outra parte desse trabalho foi focada no uso de imagens multiespectrais (TM/Landsat-5 e OLI/Landsat-8) e hiperespectrais (Hyperion/EO-1). Usando o limiar 0,53 da imagem fraÃÃo solo obtida de um modelo de mistura espectral aplicado sobre os dados do sensor OLI (Setembro/2013) e informaÃÃes do comportamento temporal (1984-2011) do Ãndice de vegetaÃÃo por diferenÃa normalizada (NDVI) obtido do sensor TM, Ãreas de solo exposto foram avaliadas quanto à sua diferenciaÃÃo nas classes "salinizados" e "nÃo-salinizados". Na discriminaÃÃo desses alvos tambÃm foram usados Ãndices de salinidade e ACP obtidos de dados dos sensores OLI e Hyperion. Com dados desses dois sensores, tambÃm foi averiguada a capacidade de Ãndices multiespectrais e hiperespectrais de vegetaÃÃo em identificar e caracterizar o estresse salino em dossÃis de arroz. Foram usadas regressÃes lineares para descrever a relaÃÃo entre os Ãndices e a CE do solo. A espectroscopia de laboratÃrio revelou que as amostras NTG apresentaram uma diminuiÃÃo na reflectÃncia e brilho com a salinizaÃÃo usando CaCl2 e MgCl2 e um aumento usando NaCl. O gesso aumentou a reflectÃncia do solo e foi determinante para a apariÃÃo da banda de absorÃÃo em 1750 nm nos espectros das amostras TG. As bandas de absorÃÃo mais importantes verificadas nos espectros salinizados foram observadas em 1450, 1950 e 1750 nm. O modelo preditivo desenvolvido com NDSI (R2 = 0,84), a partir de bandas do espectrÃmetro posicionadas prÃximas a 1900 nm, apresentaram resultados superiores aos modelos de reflectÃncia de bandas individuais (R2 = 0,50). No entanto, foi o PLSR (R2 = 0,88) usando todas as bandas espectrais do espectrÃmetro que apresentou os melhores resultados da modelagem sugerindo que, quanto maior o nÃmero de informaÃÃes espectrais usadas, maior à a capacidade de previsÃo dosmodelos. Com os dados do OLI foram observadas boas correlaÃÃes do Salinity Index (SI) (r = +0,84) e da primeira componente principal (CP1) (r = +0,83) com a CE do solo. Uma forte correlaÃÃo (r = +0,77) tambÃm foi observada a partir da CP1 dos dados Hyperion. Em condiÃÃes de campo, os espectros de reflectÃncia e a ACP indicaram que Ãreas com maiores CE possuem maior brilho em relaÃÃo Ãs demais Ãreas nÃo-salinizadas e isso possibilita o uso de dados dos dois sensores para discriminar solos expostos salinizados de nÃo salinizados. Para dossÃis de arroz, a reflectÃncia no infravermelho prÃximo (NIR) e infravermelho mÃdio (SWIR) foi reduzida com o aumento da CE do solo. Jà na regiÃo do vermelho, o estresse salino provocou um aumento de reflectÃncia. Isso favoreceu aos bons resultados apresentados pelo NDVI (R2 = 0,68) e Enhanced Vegetation Index (EVI) (R2 = 0,70) obtidos do sensor OLI para caracterizar a resposta espectral do arroz sob diferentes CE do solo. Os Ãndices hiperespectrais mais promissores foram o Salinity and Water Stress Index (SWSI1) (R2=0,70) e Ãndice do Estresse Salino para Arroz (IESA) (R2=0,59), que sÃo combinaÃÃes de faixas espectrais relacionadas à clorofila e à absorÃÃo de Ãgua e/ou estresse hÃdrico. No geral, o estudo mostrou que o SR tem um bom potencial de aplicaÃÃo para detectar e caracterizar Ãreas salinizadas. O uso de imagens à bastante promissor, porÃm informaÃÃes obtidas com espectroscopia de laboratÃrio sÃo necessÃrias para subsidiar o entendimento das particularidades de caracterÃsticas espectrais dos alvos. / The characterization, delineation and assessment of areas affected by salt/sodium is extremely important for the Irrigation Perimeter of Morada Nova â Cearà and can contribute in decision-making processes on local farming. Remote sensing (RS) is an attractive alternative to traditional methods of soil salinization studies due to its low cost, spatial coverage and temporal frequency of image acquisition. It may provide a fast and non-destructive mapping of the salinized areas. This study aimed to use RS data in the development of methodological strategies to identify areas with salinity problems allowing a preliminary assessment of salt effects on soil and vegetation. Initially, we used laboratory spectroscopy to characterize and quantify variations in reflectance and spectral absorption bands as function of the changes in electrical conductivity (EC) of the soils. Neossolos samples (n = 180) were salinized in laboratory with increasing concentrations of NaCl, MgCl2 and CaCl2. Half of them were previously treated with gypsum. Reflectance spectra were measured at nadir viewing in a controlled laboratory environment using the FieldSpec spectrometer. Variations in reflectance and absorption bands attributes were evaluated by using principal component analysis (PCA) and the continuum removal (CR) technique, respectively, for soils treated with gypsum (TG) and non-treated with gypsum (NTG). Using soil samples of NTG (n = 62) and a set of independent samples (n = 32) collected from various sites within the irrigated perimeter, predictive models were developed using linear regressions of individual bands, the normalized salinity index (NDSI) and partial least squares regression (PLSR). Another part of this work was focused on the use of multispectral images (TM/Landsat-5 and OLI/Landsat-8) and hyperspectral (Hyperion/EO-1). Using the 0.53 threshold over the soil fraction image from the spectral mixture model applied to the OLI data (September, 2013) and information on the temporal behavior (1984-2011) of the Normalized Difference Vegetation Index (NDVI) obtained from the TM sensor, exposed soils were evaluated for their differentiation in the saline and non-saline classes. For the discrimination of these classes and salinity levels, PCA was applied to OLI and Hyperion data. By using data from these two sensors, the ability of multispectral and hyperspectral vegetation indices to identify and evaluate salt stress in rice canopies was investigated. Linear regressions were used to describe the relationship between the indices and soil EC. Results from laboratory reflectance spectroscopy showed that NTG samples presented a decrease in reflectance and brightness after salinization with CaCl2 and MgCl2, and an increase of them after salinization with NaCl. Gypsum increased the soil reflectance and was crucial to the appearance of the absorption band at 1750 nm in the TG samples. The most important spectral features were observed in salinized spectra at 1450, 1950 and 1750 nm. The predictive model developed with NDSI (R2 = 0.836) from bands positioned close to 1900 nm showed the best results when individual bands were considered in the analysis (R2 = 0.50). However, PLSR (R2 = 0.883) using all the spectral bands showed the best model suggesting that the greatest number of bands produced the largest predictive power for the models. Using information from the OLI, statistically significant correlations of the Salinity Index (SI) (r = 0.84) and first principal component (PC1) (r = 0.83) with the soil EC were obtained. A significant correlation (r = 0.77) was also observed with the PC1 of Hyperion data. Under field conditions, the spectral profiles and PCA indicated that areas with higher EC had also greater brightness relative to the non-salinized areas, which enabled the use of data from the two sensors to discriminate the exposed salinized soils from the non-salinized ones. For rice, canopy reflectance in the near infrared (NIR) and shortwave infrared (SWIR) was reduced with increasing soil EC. In the red spectral region of chlorophyll absorption, the salt stress caused a slightly reflectance increase. This explained the good results presented by NDVI (R2 = 0.68) and Enhanced Vegetation Index (EVI) (R2 = 0.70) obtained from the OLI sensor to characterize the spectral response of rice under different soil ECs. The most promising hyperspectral indices were the Salinity and Water Stress Index (SWSI1) (R2 = 0.70) and Saline Stress Index for Rice (IESA) (R2 = 0.59), which are combinations of regions related to chlorophyll regions with absorption of water that vary with water stress. Overall, this study showed that the RS has a good potential to detect and characterize salinization areas. The use of images is very promising, but information obtained from laboratory spectroscopy provides the necessary understanding of the particularities of spectral characteristics of the saline soils.

La disponibilité de nouveaux moyens d’acquisition en télédétection, satellitaire (PLEIADES, HYPXIM), aéroportée ou par drone (UAV) à très haute résolution spatiale ouvre la voie à leur utilisation pour l’étude de milieux complexes telles que les villes. En particulier, la connaissance de la ville pour l’étude des îlots de chaleur, la planification urbaine, l’estimation de la biodiversité de la végétation et son état de santé nécessite au préalable une étape de classification des matériaux qui repose sur l’utilisation de l’information spectrale accessible en télédétection hyperspectrale 0,4-2,5μm. Une des principales limitations des méthodes de classification réside dans le non traitement des zones à l’ombre. Des premiers travaux ont montré qu’il était possible d’exploiter l’information radiative dans les ombres des bâtiments. En revanche, les méthodes actuelles ne fonctionnent pas dans les ombres des arbres du fait de la porosité de leur couronne. L’objectif de cette thèse vise à caractériser les propriétés optiques de surface à l’ombre de la végétation arborée urbaine au moyen d’outils de transfert radiatif et de correction atmosphérique. L’originalité de ce travail est d’étudier la porosité d’un arbre via la grandeur de transmittance de la couronne. La problématique a donc été abordée en deux temps. Premièrement, la caractérisation de la transmittance d’un arbre isolé a été menée avec l’utilisation de l’outil DART à travers la mise en œuvre d’un plan d’expériences et d’études de sensibilité qui ont permis de la relier à des paramètres biophysiques et externes. Une campagne de mesures terrain a ensuite été réalisée afin d’évaluer son estimation à partir de différents niveaux de modélisation de l’arbre, dont un modèle réel acquis par mesures lidar terrestre. Deuxièmement, une nouvelle méthode de correction atmosphérique 3D adaptée à la végétation urbaine, ICARE-VEG, a été développée à partir des résultats précédents. Une campagne aéroportée et de mesures terrain UMBRA a été dédiée à sa validation. Ses performances comparées à d’autres outils existants ouvrent de larges perspectives pour l’interprétation globale d’une image par télédétection et pour souligner la complexité de modéliser des processus physiques naturels à une échelle spatiale très fine. / The new advances in remote sensing acquisitions at very high spatial resolution, either spaceborne (PLEIADES, HYPXIM), airborne or unmanned aerial vehicles borne, open the way for the study of complex environments such as urban areas. In particular, the better understanding of urban heat islands, urban planning, vegetation biodiversity, requires the knowledge of detailed material classification mapsbased on the use of spectral information brought by hyperspectral imagery 0.4-2.5μm. However, one of the main limitations of classification methods relies on the absence of shadow processing. Past studies have demonstrated that spectral information was possible to be extracted from shadows cast by buildings. But existing methods fail in shadows cast by trees because of their crown porosity. The objective of this thesis aims to characterize surface optical properties in urban tree shadows by means of radiative transfer and atmospheric correction tools. The originality of this work is to study the tree crown porosity through the analysis of the tree crown transmittance. Therefore, the issue has been divided into two parts. Firstly, an experimental design with the use of DART tool has been carried out in order to examine the relationships between the transmittance of an isolated tree and different biophysical and external variables. Then, the estimation of the tree crown transmittance has been assessed with several tree 3D modelling strategies derived from reference terrestrial lidar acquisitions. Secondly, a new atmospheric correction method appropriate to the processing of tree shadows, ICARE-VEG, was implemented fromthese previous results. An airborne and field campaign UMBRA was dedicated to its validation. Moreover, its performances was compared to other existing tools. Finally, the conclusions open large outlooks to the overall interpretation of remote sensing images and highlight the complexity to model physical natural processes with finer spatial resolutions.