Résumé non communiqué par le doctorant. / The optimal exploitation of the information provided by hyperspectral images requires the development of advanced image processing tools. Therefore, under the title Hyperspectral image representation and Processing with Binary Partition Trees, this PhD thesis proposes the construction and the processing of a new region-based hierarchical hyperspectral image representation:the Binary Partition Tree (BPT). This hierarchical region-based representation can be interpretedas a set of hierarchical regions stored in a tree structure. Hence, the Binary Partition Tree succeedsin presenting: (i) the decomposition of the image in terms of coherent regions and (ii) the inclusionrelations of the regions in the scene. Based on region-merging techniques, the construction of BPTis investigated in this work by studying hyperspectral region models and the associated similaritymetrics. As a matter of fact, the very high dimensionality and the complexity of the data require the definition of specific region models and similarity measures. Once the BPT is constructed, the fixed tree structure allows implementing efficient and advanced application-dependent techniqueson it. The application-dependent processing of BPT is generally implemented through aspecific pruning of the tree. Accordingly, some pruning techniques are proposed and discussed according to different applications. This Ph.D is focused in particular on segmentation, object detectionand classification of hyperspectral imagery. Experimental results on various hyperspectraldata sets demonstrate the interest and the good performances of the BPT representation.

<div>Maize (Zea mays L.) is one of the most important crops worldwide for its critical importance in agriculture, economic stability, and food security. Many agricultural research and commercial breeding programs target the efficiency of this crop, seeking to increase productivity with fewer inputs and becoming more environmentally sustainable and resistant to impacts of climate and other external factors. For the purpose of analyzing the performance of the new varieties and management strategies, accurate and constant monitoring is crucial and yet, still performed mostly manually, becoming labor-intensive, time-consuming, and costly.<br></div><div>Flowering is one of the most important stages for maize, and many other grain crops, requiring close attention during this period. Any physical or biological negative impact in the tassel, as a reproductive organ, can have significant consequences to the overall grain development, resulting in production losses. Remote sensing observation technologies are currently seeking to close the gap in phenotyping in monitoring the development of the plants’ geometric structure and chemistry-related responses over the growth and reproductive cycle.</div><div>For this thesis, remotely sensed hyperspectral imagery were collected, processed and, explored to detect tassels in maize crops. The data were acquired in both a controlled facility using an imaging conveyor, and from the fields using a PhenoRover (wheel-based platform) and a low altitude UAV. Two pixel-based classification experiments were performed on the original hyperspectral imagery (HSI) using Spectral Angle Mapper (SAM) and Support Vector Machine (SVM) supervised classifiers. Feature reduction methods, including Principal Component Analysis (PCA), Locally Linear Embedding (LLE), and Isometric Feature Mapping (Isomap) were also investigated, both to identify features for annotating the reference data and in conjunction with classification.</div><div>Collecting the data from different systems allowed the identification of strengths and weaknesses for each system and the associated tradeoffs. The controlled facility allowed stable lighting and very high spatial and spectral resolution, although it lacks on supplying information about the plants’ interactions in field conditions. Contrarily, the in-field data from the PhenoRover </div><div>and the UAV exposed the complications related to the plant’s density within the plots and the variability in the lighting conditions due to long times of data collection required. The experiments implemented in this study successfully classified pixels as tassels for all images, performing better with higher spatial resolution and in the controlled environment. For the SAM experiment, nonlinear feature extraction via Isomap was necessary to achieve good results, although at a significant computational expense. Dimension reduction did not improve results for the SVM classifier.</div>

No description available.

Flue-cured tobacco (Nicotiana tabacum L.) is a high value-per-acre crop that is intensively managed to optimize the yield of high quality cured leaf. Aerial remote sensing, specifically unmanned aerial vehicles (UAVs), present flue-cured tobacco producers and researchers with a potential tool for scouting and crop management. A two-year study, conducted in Southside Virginia at the Southern Piedmont Agricultural Research and Extension Center and on commercial farms, assessed the potential of aerial remote sensing in flue-cured tobacco. The effort encompassed two key objectives. First, examine the use of the enhanced normalized difference vegetation index (ENDVI) for separating flue-cured tobacco varieties and nitrogen rates. Secondly, develop hyperspectral indices and/or machine learning classification models capable of detecting Phytophthora nicotianae (black shank) incidence in flue-cured tobacco. In 2017, UAV-acquired ENDVI surveys demonstrated the ability to consistently separate between flue-cured tobacco varieties and nitrogen rates from topping to harvest. In 2018, ENDVI revealed significant differences among N-rates as early as 34 days after transplanting. Two hyperspectral indices were developed to detect black shank incidence based on differences in the spectral profiles of asymptomatic flue-cured tobacco plants compared to those with black shank symptoms. Testing of the indices showed significant differences between the index values of healthy and symptomatic plants (alpha = 0.05). In addition, the indices were able to detect black shank symptoms pre-symptomatically (alpha = 0.09). Subspace linear discriminant analysis, a machine learning classification, was also used for prediction of black shank incidence with up to 85.7% classification accuracy. / Master of Science / Unmanned Aerial Vehicle’s (UAVs) or drones, as they are commonly referred to, may have potential as a tool in flue-cured tobacco research and production. UAVs combined with sensors and cameras provide the opportunity to gather a large amount of data on a particular crop, which may be useful in crop management. Given the intensive management of flue-cured tobacco, producers may benefit from extra insight on how to better assess threats to yield such as under-fertilization and disease pressure. A two-year study was conducted in Southside Virginia at the Southern Piedmont Agricultural Research and Extension Center and on commercial farms. There were two objectives to this effort. First, assess the ability of UAV-acquired multispectral near-infrared imagery to separate flue-cured tobacco varieties and nitrogen rates. Secondly, develop hyperspectral indices and machine learning models that can accurately predict the incidence of black shank in flue-cured tobacco. Flue-cured tobacco nitrogen rates were significantly different in 2017 from 59 days after transplanting to harvest using UAV-acquired near-infrared imagery. In 2018, heavy rainfall may have led to nitrogen leaching from the soil resulting in nitrogen rates being significantly different as early as 34 days after transplanting. The imagery also showed a significant relationship with variety maturation type in the late stages of crop development during ripening. Two hyperspectral indices were developed and one machine learning model was trained. Each had the ability to detect black shank incidence in fluecured tobacco pre-symptomatically, as well as separated black shank infested plants from healthy plants.

L’imagerie hyperspectrale est très largement employée en télédétection pour diverses applications, dans le domaine civil comme dans le domaine militaire. Une image hyperspectrale est le résultat de l’acquisition d’une seule scène observée dans plusieurs longueurs d’ondes. Par conséquent, chacun des pixels constituant cette image est représenté par un vecteur de mesures (généralement des réflectances) appelé spectre. Une étape majeure dans l’analyse des données hyperspectrales consiste à identifier les composants macroscopiques (signatures) présents dans la région observée et leurs proportions correspondantes (abondances). Les dernières techniques développées pour ces analyses ne modélisent pas correctement ces images. En effet, habituellement ces techniques supposent l’existence de pixels purs dans l’image, c’est-à-dire des pixels constitué d’un seul matériau pur. Or, un pixel est rarement constitué d’éléments purs distincts l’un de l’autre. Ainsi, les estimations basées sur ces modèles peuvent tout à fait s’avérer bien loin de la réalité. Le but de cette étude est de proposer de nouveaux algorithmes d’estimation à l’aide d’un modèle plus adapté aux propriétés intrinsèques des images hyperspectrales. Les paramètres inconnus du modèle sont ainsi déduits dans un cadre Bayésien. L’utilisation de méthodes de Monte Carlo par Chaînes de Markov (MCMC) permet de surmonter les difficultés liées aux calculs complexes de ces méthodes d’estimation. / Hyperspectral imagery has been widely used in remote sensing for various civilian and military applications. A hyperspectral image is acquired when a same scene is observed at different wavelengths. Consequently, each pixel of such image is represented as a vector of measurements (reflectances) called spectrum. One major step in the analysis of hyperspectral data consists of identifying the macroscopic components (signatures) that are present in the sensored scene and the corresponding proportions (concentrations). The latest techniques developed for this analysis do not properly model these images. Indeed, these techniques usually assume the existence of pure pixels in the image, i.e. pixels containing a single pure material. However, a pixel is rarely composed of pure spectrally elements, distinct from each other. Thus, such models could lead to weak estimation performance. The aim of this thesis is to propose new estimation algorithms with the help of a model that is better suited to the intrinsic properties of hyperspectral images. The unknown model parameters are then infered within a Bayesian framework. The use of Markov Chain Monte Carlo (MCMC) methods allows one to overcome the difficulties related to the computational complexity of these inference methods.

A fundamental challenge in target detection in hyperspectral imagery is spectral variability. In target detection applications, we are provided with a pure target signature; we do not have a collection of samples that characterize the spectral variability of the target. Another problem is that the performance of stochastic detection algorithms such as the spectral matched filter can be detrimentally affected by the assumptions of multivariate normality of the data, which are often violated in practical situations. We address the challenge of lack of training samples by creating two models to characterize the target class spectral variability --the first model makes no assumptions regarding inter-band correlation, while the second model uses a first-order Markovbased scheme to exploit correlation between bands. Using these models, we present two techniques for meeting these challenges-the kernel-based support vector data description (SVDD) and spectral fringe-adjusted joint transform correlation (SFJTC). We have developed an algorithm that uses the kernel-based SVDD for use in full-pixel target detection scenarios. We have addressed optimization of the SVDD kernel-width parameter using the golden-section search algorithm for unconstrained optimization. We investigated a proper number of signatures N to generate for the SVDD target class and found that only a small number of training samples is required relative to the dimensionality (number of bands). We have extended decision-level fusion techniques using the majority vote rule for the purpose of alleviating the problem of selecting a proper value of s 2 for either of our target variability models. We have shown that heavy spectral variability may cause SFJTC-based detection to suffer and have addressed this by developing an algorithm that selects an optimal combination of the discrete wavelet transform (DWT) coefficients of the signatures for use as features for detection. For most scenarios, our results show that our SVDD-based detection scheme provides low false positive rates while maintaining higher true positive rates than popular stochastic detection algorithms. Our results also show that our SFJTC-based detection scheme using the DWT coefficients can yield significant detection improvement compared to use of SFJTC using the original signatures and traditional stochastic and deterministic algorithms.

Les émissions atmosphériques constitue un enjeu majeur pour la société, à la fois pour les problématiques santé – qualité de l’air (maladies respiratoires, allergies,. . . ) et pour les problématiques liées au réchauffement climatique et aux gaz à effet de serre. Les sources anthropiques, industrielles en particulier, émettent dans l’atmosphère gaz et aérosols qui jouent un rôle important dans les échanges atmosphériques. Néanmoins leur suivi à haute résolution spatiale reste peu précis, en raison des résolutions rencontrées pour les senseurs spatiaux. Les développements techniques récents des capteurs hyperspectraux aéroportés permettent d’améliorer la caractérisation des panaches. Lors de nos travaux nous avons développé un nouvel outil pour la détection et la caractérisation des panaches de gaz. Ainsi, une cartographie 3D des différentes concentrations est obtenue. Cet outil est ensuite validé sur des images synthétiques et sur des acquisitions aéroportés de scènes industrielles. / The air pollution is a very important issue for industrialized society, both in terms of health (respiratory diseases, allergies,. . . ) and in terms of climate change (global warming and greenhouse gas emissions). Anthropogenic sources, especially industrial, emit into the atmosphere gases and aerosols, which play an important role in atmospheric exchanges. However emissions remain poorly estimated as most of existing space sensors have a limited spectral range as well as a too low spatial resolution. The use of the new hyperspectral airborne image sensors in the infrared range opens the way to new development to improve the plume characterization. In our work, we developed a new method for detecting and characterizing gas plumes. It is based on an accurate non linear formalism of cloud gas radiative impact. This method was validated using synthetic scenes of industrial area, and airborne acquisitions obtained by a hyperspectral thermal infrared sensor.

Cette thèse vise à proposer de nouveaux modèles pour la séparation de sources dans le cadre non linéaire des méthodes à noyaux en apprentissage statistique, et à développer des algorithmes associés. Le domaine d'application privilégié est le démélange en imagerie hyperspectrale. Tout d'abord, nous décrivons un modèle original de la factorisation en matrices non négatives (NMF), en se basant sur les méthodes à noyaux. Le modèle proposé surmonte la malédiction de préimage, un problème inverse hérité des méthodes à noyaux. Dans le même cadre proposé, plusieurs extensions sont développées pour intégrer les principales contraintes soulevées par les images hyperspectrales. Pour traiter des masses de données, des algorithmes de traitement en ligne sont développés afin d'assurer une complexité calculatoire fixée. Également, nous proposons une approche de factorisation bi-objective qui permet de combiner les modèles de démélange linéaire et non linéaire, où les décompositions de NMF conventionnelle et à noyaux sont réalisées simultanément. La dernière partie se concentre sur le démélange robuste aux bandes spectrales aberrantes. En décrivant le démélange selon le principe de la maximisation de la correntropie, deux problèmes de démélange robuste sont traités sous différentes contraintes soulevées par le problème de démélange hyperspectral. Des algorithmes de type directions alternées sont utilisés pour résoudre les problèmes d'optimisation associés / This thesis aims to propose new nonlinear unmixing models within the framework of kernel methods and to develop associated algorithms, in order to address the hyperspectral unmixing problem.First, we investigate a novel kernel-based nonnegative matrix factorization (NMF) model, that circumvents the pre-image problem inherited from the kernel machines. Within the proposed framework, several extensions are developed to incorporate common constraints raised in hypersepctral images analysis. In order to tackle large-scale and streaming data, we next extend the kernel-based NMF to an online fashion, by keeping a fixed and tractable complexity. Moreover, we propose a bi-objective NMF model as an attempt to combine the linear and nonlinear unmixing models. The decompositions of both the conventional NMF and the kernel-based NMF are performed simultaneously. The last part of this thesis studies a supervised unmixing model, based on the correntropy maximization principle. This model is shown robust to outlier bands. Two correntropy-based unmixing problems are addressed, considering different constraints in hyperspectral unmixing problem. The alternating direction method of multipliers (ADMM) is investigated to solve the related optimization problems

This dissertation develops several new algorithms to solve existing problems in practical application of the previously developed PCA+JPEG2000, which has shown superior rate-distortion performance in hyperspectral image compression. In addition, a new scheme is proposed to facilitate multi-temporal hyperspectral image compression. Specifically, the uniqueness in each algorithm is described as follows. 1. An empirical piecewise linear equation is proposed to estimate the optimal number of major principal components (PCs) used in SubPCA+JPEG2000 for AVIRIS data. Sensor-specific equations are presented with excellent fitting performance for AVIRIS, HYDICE, and HyMap data. As a conclusion, a general guideline is provided for finding sensor-specific piecewise linear equations. 2. An anomaly-removal-based hyperspectral image compression algorithm is proposed. It preserves anomalous pixels in a lossless manner, and yields the same or even improved rate-distortion performance. It is particularly useful to SubPCA+JPEG2000 when compressing data with anomalies that may reside in minor PCs. 3. A segmented PCA-based PCA+JPEG2000 compression algorithm is developed, which spectrally partitions an image based on its spectral correlation coefficients. This compression scheme greatly improves the rate-distortion performance of PCA+JPEG2000 when the spatial size of the data is relatively smaller than its spectral size, especially at low bitrates. A sensor-specific partition method is also developed for fast processing with suboptimal performance. 4. A joint multi-temporal image compression scheme is proposed. The algorithm preserves change information in a lossless fashion during the compression. It can yield perfect change detection with slightly degraded rate-distortion performance.

No description available.