The objective of this dissertation is to investigate all the necessary components in spectral mixture analysis (SMA) for hyperspectral imagery under an unsupervised circumstance. When SMA is linear, referred to as linear spectral mixture analysis (LSMA), these components include estimation of the number of endmembers, extraction of endmember signatures, and calculation of endmember abundances that can automatically satisfy the sum-to-one and non-negativity constraints. A simple approach for nonlinear spectral mixture analysis (NLSMA) is also investigated. After SMA is completed, a color display is generated to present endmember distribution in the image scene. It is expected that this research will result in an analytic system that can yield optimal or suboptimal SMA output without prior information. Specifically, the uniqueness in each component is described as follow. 1)A new signal subspace-based approach is developed to determine the number of endmembers with relatively robust performance and the least parameter requirement. 2)The best implementation strategy is determined for endmember extraction algorithms using simplex volume maximization and pixel spectral similarity; and algorithm with the special consideration for anomalous pixels is developed to improve the quality of extracted endmembers. 3)A new linear mixture model (LMM) is deployed where the number of endmembers and their types can be changed from pixel to pixel such that the resulting endmember abundances are sum-to-one and nonnegative as required; and fast algorithms are developed to search for a sub-optimal endmember set for each pixel. 4)A simple approach for NLSMA based on LMM is investigated and an automated approach is developed to determine either linear or nonlinear mixing is actually experienced. 5)A color display strategy is developed to present SMA results with high class/endmember separability.

A Hyperspectral imager aboard a cubesat would be able to provide images which could be used to identify ships and determine the ship's length and breadth and heading. Depending on the size of the ship, the speed the ship is traveling can be determined as well; however the speed and size determination is limited by the spatial resolution of 100 meters. The spectral signature of the boat is dramatically different from the spectral signature of the open Ocean especially within the range of 400 to 1000 nanometers, and this threshold is the basis for extracting ship data. Hyperspectral Imagers are ideal for minimization with few optical errors introduced, and designs range in durability making them useful on board small satellites especially in the visible and near infrared region. Placing an imager on a satellite allows for consistent observation over a region to identify patterns in ship movement over time. / Master of Science

Hyperspectral imaging (HSI) produces spatial images with pixels that, instead of consisting of three colors, consist of hundreds of spectral measurements. Because there are so many measurements for each pixel, analysis of HSI is difficult. Frequently, standard techniques are used to help make analysis more tractable by representing the HSI data in a different manner. This research explores the utility of representing the HSI data in a learned dictionary basis for the express purpose of material identification and classification. Multiclass classification is performed on the transformed data using the RandomForests algorithm. Performance results are reported. In addition to classification, single material detection is considered also. Commonly used detection algorithm performance is demonstrated on both raw radiance pixels and HSI represented in dictionary-learned bases. Comparison results are shown which indicate that detection on dictionary-learned sparse representations perform as well as detection on radiance. In addition, a different method of performing detection, capitalizing on dictionary learning is established and performance comparisons are reported, showing gains over traditional detection methods.

The debris-flow initiation variable of clay mineralogy is examined for Holocene age debris-flow deposits across the Colorado Plateau. A Kolmogorov-Smirnov two-sample test between 25 debris-flow producing shale units and 23 shale units rated as not producing debris-flows found a highly significant difference between shale unit kaolinite-illite and montmorillonite clay content. Debris-flow producers tend to have abundant kaolinite and illite (61.5% of clays) and small amounts of montmorillonite (10.4%). Clay sample soluble cation (Na, Ca, K, and Mg) content could not be used to accurately divide the data set into debris-flow producers and debris-flow non-producers by either cluster analysis or a Kolmogorov-Smirnov two-sample test.AVIRIS hyperspectral data reveal that debris-flow deposits, colluvium, and some shale units in Cataract Canyon, Utah display the double-absorption feature characteristic of kaolinite at 2.2 µm. Lab-based reflection spectra and semi-quantitative x-ray diffraction results show that Cataract Canyon debris-flow matrix clays are dominated by kaolinite and illite and lacking in montmorillonite. A surface material map showing the spectral stratigraphy of the study area was created from AVIRIS data classified using an artificial neural network and compares favorably to existing geologic data for Cataract Canyon. A debris-flow initiation potential map created from a GIS-based analysis of surface materials, slope steepness, slope aspect, and fault maps shows the greatest debris-flow initiation potential in the study area to coincide with outcrops of the Moenkopi Formation on steep (>20%), southwest-facing slopes. Small areas of extreme debris-flow initiation potential are located where kaolinite and illite clay-rich colluvial wedges are located on southwest-facing walls of Colorado River tributary canyons. The surface materials map shows formations clearly when they remain relatively consistent in composition and exposure throughout the study area, such as the White Rim Sandstone and most clay-rich members of the Moenkopi Formation. The debris-flow producing Organ Rock Shale and Halgaito Formation were shown inconsistently on the surface materials map, likely as a result of compositional variations in the study area. The results of this study provides evidence that hyperspectral imagery classified using an ANN can be successfully used to map the spectral stratigraphy of a sparsely vegetated area such as Cataract Canyon.

Agriculture faces many challenges related to the increasing food demands of a growing global population and the sustainable use of resources in a changing environment. To address them, we need reliable information sources, like exploiting hyperspectral satellite, airborne, and ground-based remote sensing data to observe phenological traits through a crops growth cycle and gather information to precisely diagnose when, why, and where a crop is suffering negative impacts. By combining hyperspectral capabilities with unmanned aerial vehicles (UAVs), there is an increased capacity for providing time-critical monitoring and new insights into patterns of crop development. However, considerable effort is required to effectively utilize UAV-integrated hyperspectral systems in crop-modeling and crop-breeding tasks. Here, a UAV-based hyperspectral solution for mapping crop physiological parameters was explored within a machine learning framework. To do this, a range of complementary measurements were collected from a field-based phenotyping experiment, based on a diversity panel of wild tomato (Solanum pimpinellifolium) that were grown under fresh and saline conditions. From the UAV data, positionally accurate reflectance retrievals were produced using a computationally robust automated georectification and mosaicking methodology. The resulting multitemporal UAV data were then employed to retrieve leaf-chlorophyll (Chl) dynamics via a machine learning framework. Several approaches were evaluated to identify the best-performing regression supervised methods. An investigation of two learning strategies (i.e., sequential and retraining) and the value of using spectral bands and vegetation indices (VIs) as prediction features was also performed. Finally, the utility of UAVbased hyperspectral phenotyping was demonstrated by detecting the effects of salt-stress on the different tomato accessions by estimating the salt-induced senescence index from the retrieved Chl dynamics, facilitating the identification of salt-tolerant candidates for future investigations. This research illustrates the potential of UAV-based hyperspectral imaging for plant phenotyping and precision agriculture. In particular, a) developing systematic imaging calibration and pre-processing workflows; b) exploring machine learning-driven tools for retrieving plant phenological dynamics; c) establishing a plant stress detection approach from hyperspectral-derived metrics; and d) providing new insights into using computer vision, big-data analytics, and modeling strategies to deal effectively with the complexity of the UAV-based hyperspectral data in mapping plant physiological indicators.

Processing hyperspectral image data can be computationally expensive and difficult to employ for real-time applications due to its extensive spatial and spectral information. Further, applications in which computational resources may be limited can be hindered by the volume of data that is common with airborne hyperspectral image data. This paper proposes utilizing band selection to down-select the number of spectral bands to consider for a given classification task such that classification can be done at the edge. Specifically, we consider the following state of the art band selection techniques: Fast Volume-Gradient-based Band Selection (VGBS), Improved Sparse Subspace Clustering (ISSC), Maximum-Variance Principal Component Analysis (MVPCA), and Normalized Cut Optimal Clustering MVPCA (NC-OC-MVPCA), to investigate their feasibility at identifying discriminative bands such that classification performance is not drastically hindered. This would greatly benefit applications where time-sensitive solutions are needed to ensure optimal outcomes. In this research, an NVIDIA AGX Xavier module is used as the edge device to run trained models on as a simulated deployed unmanned aerial system. Performance of the proposed approach is measured in terms of classification accuracy and run time.

Airborne multi-sensor imaging is increasingly used to examine vegetation properties. The advantage of using multiple types of sensor is that each detects a different feature of the vegetation, so that collectively they provide a detailed understanding of the ecological pattern. Specifically, Light Detection And Ranging (LiDAR) devices produce detailed point clouds of where laser pulses have been backscattered from surfaces, giving information on vegetation structure; hyperspectral sensors measure reflectances within narrow wavebands, providing spectrally detailed information about the optical properties of targets; while aerial photographs provide high spatial-resolution imagery so that they can provide more feature details which cannot be identified from hyperspectral or LiDAR intensity images. Using a combination of these sensors, effective techniques can be developed for mapping species and inferring leaf physiological processes at ITC-level. Although multi-sensor approaches have revolutionised ecological research, their application in mapping individual tree crowns is limited by two major technical issues: (a) Multi-sensor imaging requires all images taken from different sensors to be co-aligned, but different sensor characteristics result in scale, rotation or translation mismatches between the images, making correction a pre-requisite of individual tree crown mapping; (b) reconstructing individual tree crowns from unstructured raw data space requires an accurate tree delineation algorithm. This thesis develops a schematic way to resolve these technical issues using the-state-of-the-art computer vision algorithms. A variational method, called NGF-Curv, was developed to co-align hyperspectral imagery, LiDAR and aerial photographs. NGF-Curv algorithm can deal with very complex topographic and lens distortions efficiently, thus improving the accuracy of co-alignment compared to established image registration methods for airborne data. A graph cut method, named MCNCP-RNC was developed to reconstruct individual tree crowns from fully integrated multi-sensor imagery. MCNCP-RNC is not influenced by interpolation artefacts because it detects trees in 3D, and it detects individual tree crowns using both hyperspectral imagery and LiDAR. Based on these algorithms, we developed a new workflow to detect species at pixel and ITC levels in a temperate deciduous forest in the UK. In addition, we modified the workflow to monitor physiological responses of two oak species with respect to environmental gradients in a Mediterranean woodland in Spain. The results show that our scheme can detect individual tree crowns, find species and monitor physiological responses of canopy leaves.

<p>The subject of this thesis is the application of Support Vector Machines on two totally different applications, facial expressions recognition and remote sensing.</p><p>The basic idea of kernel algorithms is to transpose input data in a higher dimensional space, the feature space, in which linear operations on the data can be processed more easily. These operations in the feature space can be expressed in terms of input data thanks to the kernel functions. Support Vector Machines is a classifier using this kernel method by computing, in the feature space and on basis of examples of the different classes, hyperplanes that separate the classes. The hyperplanes in the feature space correspond to non linear surfaces in the input space.</p><p>Concerning facial expressions, the aim is to train and test a classifier able to recognise, on basis of some pictures of faces, which emotion (among these six ones: anger, disgust, fear, joy, sad, and surprise) that is expressed by the person in the picture. In this application, each picture has to be seen has a point in an N-dimensional space where N is the number of pixels in the image.</p><p>The second application is the detection of camouflage nets hidden in vegetation using a hyperspectral image taken by an aircraft. In this case the classification is computed for each pixel, represented by a vector whose elements are the different frequency bands of this pixel.</p>

The subject of this thesis is the application of Support Vector Machines on two totally different applications, facial expressions recognition and remote sensing. The basic idea of kernel algorithms is to transpose input data in a higher dimensional space, the feature space, in which linear operations on the data can be processed more easily. These operations in the feature space can be expressed in terms of input data thanks to the kernel functions. Support Vector Machines is a classifier using this kernel method by computing, in the feature space and on basis of examples of the different classes, hyperplanes that separate the classes. The hyperplanes in the feature space correspond to non linear surfaces in the input space. Concerning facial expressions, the aim is to train and test a classifier able to recognise, on basis of some pictures of faces, which emotion (among these six ones: anger, disgust, fear, joy, sad, and surprise) that is expressed by the person in the picture. In this application, each picture has to be seen has a point in an N-dimensional space where N is the number of pixels in the image. The second application is the detection of camouflage nets hidden in vegetation using a hyperspectral image taken by an aircraft. In this case the classification is computed for each pixel, represented by a vector whose elements are the different frequency bands of this pixel.

Le démélange spectral est un des sujets majeurs de l’analyse d’images hyperspectrales. Ce problème consiste à identifier les composants macroscopiques présents dans une image hyperspectrale et à quantifier les proportions (ou abondances) de ces matériaux dans tous les pixels de l’image. La plupart des algorithmes de démélange suppose un modèle de mélange linéaire qui est souvent considéré comme une approximation au premier ordre du mélange réel. Cependant, le modèle linéaire peut ne pas être adapté pour certaines images associées par exemple à des scènes engendrant des trajets multiples (forêts, zones urbaines) et des modèles non-linéaires plus complexes doivent alors être utilisés pour analyser de telles images. Le but de cette thèse est d’étudier de nouveaux modèles de mélange non-linéaires et de proposer des algorithmes associés pour l’analyse d’images hyperspectrales. Dans un premier temps, un modèle paramétrique post-non-linéaire est étudié et des algorithmes d’estimation basés sur ce modèle sont proposés. Les connaissances a priori disponibles sur les signatures spectrales des composants purs, sur les abondances et les paramètres de la non-linéarité sont exploitées à l’aide d’une approche bayesienne. Le second modèle étudié dans cette thèse est basé sur l’approximation de la variété non-linéaire contenant les données observées à l’aide de processus gaussiens. L’algorithme de démélange associé permet d’estimer la relation non-linéaire entre les abondances des matériaux et les pixels observés sans introduire explicitement les signatures spectrales des composants dans le modèle de mélange. Ces signatures spectrales sont estimées dans un second temps par prédiction à base de processus gaussiens. La prise en compte d’effets non-linéaires dans les images hyperspectrales nécessite souvent des stratégies de démélange plus complexes que celles basées sur un modèle linéaire. Comme le modèle linéaire est souvent suffisant pour approcher la plupart des mélanges réels, il est intéressant de pouvoir détecter les pixels ou les régions de l’image où ce modèle linéaire est approprié. On pourra alors, après cette détection, appliquer les algorithmes de démélange non-linéaires aux pixels nécessitant réellement l’utilisation de modèles de mélange non-linéaires. La dernière partie de ce manuscrit se concentre sur l’étude de détecteurs de non-linéarités basés sur des modèles linéaires et non-linéaires pour l’analyse d’images hyperspectrales. Les méthodes de démélange non-linéaires proposées permettent d’améliorer la caractérisation des images hyperspectrales par rapport au méthodes basées sur un modèle linéaire. Cette amélioration se traduit en particulier par une meilleure erreur de reconstruction des données. De plus, ces méthodes permettent de meilleures estimations des signatures spectrales et des abondances quand les pixels résultent de mélanges non-linéaires. Les résultats de simulations effectuées sur des données synthétiques et réelles montrent l’intérêt d’utiliser des méthodes de détection de non-linéarités pour l’analyse d’images hyperspectrales. En particulier, ces détecteurs peuvent permettre d’identifier des composants très peu représentés et de localiser des régions où les effets non-linéaires sont non-négligeables (ombres, reliefs,...). Enfin, la considération de corrélations spatiales dans les images hyperspectrales peut améliorer les performances des algorithmes de démélange non-linéaires et des détecteurs de non-linéarités. / Spectral unmixing is one the major issues arising when analyzing hyperspectral images. It consists of identifying the macroscopic materials present in a hyperspectral image and quantifying the proportions of these materials in the image pixels. Most unmixing techniques rely on a linear mixing model which is often considered as a first approximation of the actual mixtures. However, the linear model can be inaccurate for some specific images (for instance images of scenes involving multiple reflections) and more complex nonlinear models must then be considered to analyze such images. The aim of this thesis is to study new nonlinear mixing models and to propose associated algorithms to analyze hyperspectral images. First, a ost-nonlinear model is investigated and efficient unmixing algorithms based on this model are proposed. The prior knowledge about the components present in the observed image, their proportions and the nonlinearity parameters is considered using Bayesian inference. The second model considered in this work is based on the approximation of the nonlinear manifold which contains the observed pixels using Gaussian processes. The proposed algorithm estimates the relation between the observations and the unknown material proportions without explicit dependency on the material spectral signatures, which are estimated subsequentially. Considering nonlinear effects in hyperspectral images usually requires more complex unmixing strategies than those assuming linear mixtures. Since the linear mixing model is often sufficient to approximate accurately most actual mixtures, it is interesting to detect pixels or regions where the linear model is accurate. This nonlinearity detection can be applied as a pre-processing step and nonlinear unmixing strategies can then be applied only to pixels requiring the use of nonlinear models. The last part of this thesis focuses on new nonlinearity detectors based on linear and nonlinear models to identify pixels or regions where nonlinear effects occur in hyperspectral images. The proposed nonlinear unmixing algorithms improve the characterization of hyperspectral images compared to methods based on a linear model. These methods allow the reconstruction errors to be reduced. Moreover, these methods provide better spectral signature and abundance estimates when the observed pixels result from nonlinear mixtures. The simulation results conducted on synthetic and real images illustrate the advantage of using nonlinearity detectors for hyperspectral image analysis. In particular, the proposed detectors can identify components which are present in few pixels (and hardly distinguishable) and locate areas where significant nonlinear effects occur (shadow, relief, ...). Moreover, it is shown that considering spatial correlation in hyperspectral images can improve the performance of nonlinear unmixing and nonlinearity detection algorithms.