A huge effort has been made in the development of image classification systemswith the objective of creating high-quality thematic maps and to establishprecise inventories about land cover use. The peculiarities of Remote SensingImages (RSIs) combined with the traditional image classification challengesmake RSI classification a hard task. Many of the problems are related to therepresentation scale of the data, and to both the size and therepresentativeness of used training set.In this work, we addressed four research issues in order to develop effectivesolutions for interactive classification of remote sensing images.The first research issue concerns the fact that image descriptorsproposed in the literature achieve good results in various applications, butmany of them have never been used in remote sensing classification tasks.We have tested twelve descriptors that encodespectral/color properties and seven texture descriptors. We have also proposeda methodology based on the K-Nearest Neighbor (KNN) classifier for evaluationof descriptors in classification context. Experiments demonstrate that JointAuto-Correlogram (JAC), Color Bitmap, Invariant Steerable Pyramid Decomposition(SID), and Quantized Compound Change Histogram (QCCH) yield the best results incoffee and pasture recognition tasks.The second research issue refers to the problem of selecting the scaleof segmentation for object-based remote sensing classification. Recentlyproposed methods exploit features extracted from segmented objects to improvehigh-resolution image classification. However, the definition of the scale ofsegmentation is a challenging task. We have proposedtwo multiscale classification approaches based on boosting of weak classifiers.The first approach, Multiscale Classifier (MSC), builds a strongclassifier that combines features extracted from multiple scales ofsegmentation. The other, Hierarchical Multiscale Classifier (HMSC), exploits thehierarchical topology of segmented regions to improve training efficiencywithout accuracy loss when compared to the MSC. Experiments show that it isbetter to use multiple scales than use only one segmentation scale result. Wehave also analyzed and discussed about the correlation among the useddescriptors and the scales of segmentation.The third research issue concerns the selection of training examples and therefinement of classification results through multiscale segmentation. We have proposed an approach forinteractive multiscale classification of remote sensing images.It is an active learning strategy that allows the classification resultrefinement by the user along iterations. Experimentalresults show that the combination of scales produces better results thanisolated scales in a relevance feedback process. Furthermore, the interactivemethod achieves good results with few user interactions. The proposed methodneeds only a small portion of the training set to build classifiers that are asstrong as the ones generated by a supervised method that uses the whole availabletraining set.The fourth research issue refers to the problem of extracting features of ahierarchy of regions for multiscale classification. We have proposed a strategythat exploits the existing relationships among regions in a hierarchy. Thisapproach, called BoW-Propagation, exploits the bag-of-visual-word model topropagate features along multiple scales. We also extend this idea topropagate histogram-based global descriptors, the H-Propagation method. The proposedmethods speed up the feature extraction process and yield good results when compared with globallow-level extraction approaches.

L'objectif de cette thèse est de développer des solutions efficaces pour laclassification interactive des images de télédétection. Cet objectif a étéréalisé en répondant à quatre questions de recherche.La première question porte sur le fait que les descripteursd'images proposées dans la littérature obtiennent de bons résultats dansdiverses applications, mais beaucoup d'entre eux n'ont jamais été utilisés pour la classification des images de télédétection. Nous avons testé douzedescripteurs qui codent les propriétés spectrales et la couleur, ainsi que septdescripteurs de texture. Nous avons également proposé une méthodologie baséesur le classificateur KNN (K plus proches voisins) pour l'évaluation desdescripteurs dans le contexte de la classification. Les descripteurs Joint Auto-Correlogram (JAC),Color Bitmap, Invariant Steerable Pyramid Decomposition (SID) etQuantized Compound Change Histogram (QCCH), ont obtenu les meilleursrésultats dans les expériences de reconnaissance des plantations de café et depâturages.La deuxième question se rapporte au choix del'échelle de segmentation pour la classification d'images baséesur objets.Certaines méthodes récemment proposées exploitent des caractéristiques extraitesdes objets segmentés pour améliorer classification des images hauterésolution. Toutefois, le choix d'une bonne échelle de segmentation est unetâche difficile.Ainsi, nous avons proposé deux approches pour la classification multi-échelles fondées sur le les principes du Boosting, qui permet de combiner desclassifieurs faibles pour former un classifieur fort.La première approche, Multiscale Classifier (MSC), construit unclassifieur fort qui combine des caractéristiques extraites de plusieurséchelles de segmentation. L'autre, Hierarchical Multiscale Classifier(HMSC), exploite la topologie hiérarchique de régions segmentées afind'améliorer l'efficacité des classifications sans perte de précision parrapport au MSC. Les expériences montrent qu'il est préférable d'utiliser des plusieurs échelles plutôt qu'une seul échelle de segmentation. Nous avons également analysé et discuté la corrélation entre lesdescripteurs et des échelles de segmentation.La troisième question concerne la sélection des exemplesd'apprentissage et l'amélioration des résultats de classification basés sur lasegmentation multiéchelle. Nous avons proposé une approche pour laclassification interactive multi-échelles des images de télédétection. Ils'agit d'une stratégie d'apprentissage actif qui permet le raffinement desrésultats de classification par l'utilisateur. Les résultats des expériencesmontrent que la combinaison des échelles produit de meilleurs résultats que leschaque échelle isolément dans un processus de retour de pertinence. Par ailleurs,la méthode interactive permet d'obtenir de bons résultats avec peud'interactions de l'utilisateur. Il n'a besoin que d'une faible partie del'ensemble d'apprentissage pour construire des classificateurs qui sont aussiforts que ceux générés par une méthode supervisée qui utilise l'ensembled'apprentissage complet.La quatrième question se réfère au problème de l'extraction descaractéristiques d'un hiérarchie des régions pour la classificationmulti-échelles. Nous avons proposé une stratégie qui exploite les relationsexistantes entre les régions dans une hiérarchie. Cette approche, appelée BoW-Propagation, exploite le modèle de bag-of-visual-word pour propagerles caractéristiques entre les échelles de la hiérarchie. Nous avons égalementétendu cette idée pour propager des descripteurs globaux basés sur leshistogrammes, l'approche H-Propagation. Ces approches accélèrent leprocessus d'extraction et donnent de bons résultats par rapport à l'extractionde descripteurs globaux. / A huge effort has been made in the development of image classification systemswith the objective of creating high-quality thematic maps and to establishprecise inventories about land cover use. The peculiarities of Remote SensingImages (RSIs) combined with the traditional image classification challengesmake RSI classification a hard task. Many of the problems are related to therepresentation scale of the data, and to both the size and therepresentativeness of used training set.In this work, we addressed four research issues in order to develop effectivesolutions for interactive classification of remote sensing images.The first research issue concerns the fact that image descriptorsproposed in the literature achieve good results in various applications, butmany of them have never been used in remote sensing classification tasks.We have tested twelve descriptors that encodespectral/color properties and seven texture descriptors. We have also proposeda methodology based on the K-Nearest Neighbor (KNN) classifier for evaluationof descriptors in classification context. Experiments demonstrate that JointAuto-Correlogram (JAC), Color Bitmap, Invariant Steerable Pyramid Decomposition(SID), and Quantized Compound Change Histogram (QCCH) yield the best results incoffee and pasture recognition tasks.The second research issue refers to the problem of selecting the scaleof segmentation for object-based remote sensing classification. Recentlyproposed methods exploit features extracted from segmented objects to improvehigh-resolution image classification. However, the definition of the scale ofsegmentation is a challenging task. We have proposedtwo multiscale classification approaches based on boosting of weak classifiers.The first approach, Multiscale Classifier (MSC), builds a strongclassifier that combines features extracted from multiple scales ofsegmentation. The other, Hierarchical Multiscale Classifier (HMSC), exploits thehierarchical topology of segmented regions to improve training efficiencywithout accuracy loss when compared to the MSC. Experiments show that it isbetter to use multiple scales than use only one segmentation scale result. Wehave also analyzed and discussed about the correlation among the useddescriptors and the scales of segmentation.The third research issue concerns the selection of training examples and therefinement of classification results through multiscale segmentation. We have proposed an approach forinteractive multiscale classification of remote sensing images.It is an active learning strategy that allows the classification resultrefinement by the user along iterations. Experimentalresults show that the combination of scales produces better results thanisolated scales in a relevance feedback process. Furthermore, the interactivemethod achieves good results with few user interactions. The proposed methodneeds only a small portion of the training set to build classifiers that are asstrong as the ones generated by a supervised method that uses the whole availabletraining set.The fourth research issue refers to the problem of extracting features of ahierarchy of regions for multiscale classification. We have proposed a strategythat exploits the existing relationships among regions in a hierarchy. Thisapproach, called BoW-Propagation, exploits the bag-of-visual-word model topropagate features along multiple scales. We also extend this idea topropagate histogram-based global descriptors, the H-Propagation method. The proposedmethods speed up the feature extraction process and yield good results when compared with globallow-level extraction approaches.

A dissertation submitted to the Faculty of Science, University of the Witwatersrand, in fulfilment of the requirements for the Degree of Master of Science. Johannesburg, 20 January 2017. / Sand dune movements can be effectively monitored through the comparison of multitemporal satellite images. However, not all remote sensing platforms are suitable to study sand dunes. This study compares coarse (Landsat 7 and 8) and fine (Worldview 2) resolution platforms, specifically focussing on sand dunes within the Ubārī Sand Sea (Libya), and identified the average migration rate and direction for the linear dunes within a section of the Ubārī sand sea for the time period from 2002-2015 with the use of Landsat imagery. Two band combinations were compared with the use of two supervised classifications. The best combination was found to be red, green, blue and near-infrared band combination and the maximum likelihood classifier. The dune features, namely the crest, slope and interdunal areas were successfully classified based on both the coarse and fine resolution imagery, but the accuracy with which it can be classified are different between the two resolutions. The classifications based on the Worldview 2 imagery had overall accuracies ranging from 55.43 - 60.83% with kappa values of 0.3486 – 0.4225 compared to the overall accuracies and kappa values of the classifications based on the Landsat 8 imagery ranging from 52.11 – 64.67% and 0.3878 – 0.4927 respectively. An average migration rate of 8.64 (± 4.65) m/yr in a generally north western direction was calculated based on the analysis of remote sensing data with some variations in this rate and the size and shape of the dunes. It was found that although Worldview 2 imagery provides more accurate and precise mensuration data, and smaller dunes identified from Worldview data were not delineated clearly on the Landsat imagery. Landsat imagery is sufficient for the studying of dunes at a regional scale. This means that for studies concerned with the dune patterns and movements within sand seas, Landsat is sufficient. In studies where the specific dynamics of specific dunes are to be selected, a finer resolution is required; platforms such as Worldview are needed in order to gain more detailed insight and to link the past and present day climate and environmental change. / MT2017

La modélisation et la simulation de dynamiques spatiales, en particulier pour l'étude de l'évolution de paysages ou de problématiques environnementales pose la question de l'intégration des différentes formes de représentation de l'espace au sein d'un même modèle. Ocelet est une approche de modélisation de dynamiques spatiales basée sur le concept original de graphe d'interaction. Le graphe porte à la fois la structure d'une relation entre entités d’un modèle et la sémantique décrivant son évolution. Les relations entre entités spatiales sont ici traduites en graphes d'interactions et ce sont ces graphes que l'on fait évoluer lors d'une simulation. Les concepts à la base d'Ocelet peuvent potentiellement manipuler les deux formes de représentation spatiale connues, celle aux contours définis (format vecteur) ou la discrétisation en grille régulière (format raster). Le format vecteur est déjà intégré dans la première version d'Ocelet. L'intégration du format raster et la combinaison des deux restaient à étudier et à réaliser. L'objectif de la thèse est d'abord étudier les problématiques liées à l'intégration des champs continus et leur représentation discrétisée en pavage régulier, à la fois dans le langage Ocelet et dans les concepts sur lesquels il repose. Il a fallu notamment prendre en compte les aspects dynamiques de cette intégration, et d'étudier les transitions entre données géographiques de différentes formes et graphe d'interactions à l'aide de concepts formalisés. Il s'est agi ensuite de réaliser l'implémentation de ces concepts dans la plateforme de modélisation Ocelet, en adaptant à la fois son compilateur et son moteur d'exécution. Enfin, ces nouveaux concepts et outils ont été mis à l'épreuve dans trois cas d'application très différents : deux modèles sur l’île de la Réunion, le premier simulant le ruissellement dans le bassin versant de la Ravine Saint Gilles s'écoulant vers la Côte Ouest de l'île, l’autre simulant la diffusion de plantes invasives dans les plaines des hauts à l'intérieur du Parc National de La Réunion. Le dernier cas décrit la spatialisation d'un modèle de culture et est appliqué ici pour simuler les rendements de cultures céréalières sur l’ensemble de l’Afrique de l’Ouest, dans le contexte d'un système d'alerte précoce de suivi des cultures à l'échelle régionale. / The modelling and simulation of spatial dynamics, particularly for studying landscape changes or environmental issues, raises the question of integrating different forms of spatial representation within the same model. Ocelet is an approach for modelling spatial dynamics based on the original concept of interaction graph. Such a graph holds both the structure of a relation between entities of a model and the semantics describing its evolution. The relationships between spatial entities are here translated into interaction graphs and these graphs are made to evolve during a simulation. The concepts on which Ocelet is based can potentially handle two known forms of spatial representation: shapes with contours (vector format) or regular grid cells (raster). The vector format is already integrated in the first version of Ocelet. The integration of raster and the combination of the two remained to be studied and carried out. The aim of the thesis is to first study the issues related to the integration of continuous fields and their representation by regular tiling, both in the Ocelet language and the concepts on which it is based. The dynamic aspects of this integration had to be taken into account and transitions between different forms of geographic data and interaction graphs had to be studied in the light of the concepts formalized. The concepts were then implemented in the Ocelet modelling platform, with the adaptation of both its compiler and runtime. Finally, these new concepts and tools were tested in three very different cases: two models on Reunion Island, the first simulating runoff in Ravine Saint Gilles watershed in the West Coast of the island, the other simulating the spread of invasive plants in the high plains inside the Reunion National Park. The last case describes the spatialisation of a crop model and is applied here to simulate the cereal crop yields in West Africa, in the context of an early warning system for regional crop monitoring.

Orientador: Lincoln Gehring Cardoso / Banca: Luciano Nardini Gomes / Banca: Bruna Soares Xavier de Barros / Resumo: O presente trabalho objetivou comparar distâncias horizontais e áreas de um polígono considerando pontos homólogos obtidos através de levantamento topográfico convencional realizado por Estação Total Nikon Nivo 322d, levantamento georreferenciado por receptor GNSS AshTech Pro Mark 200 e imagem do Google Earth. O processamento do levantamento topográfico foi realizado através programa computacional DataGeosis versão Office que acusou elevada precisão, constituindo-se em referência. Os dados obtidos através do receptor GNSS foram pós-processados pelo software GNSS Solution e os obtidos pelo Google Earth foram submetidos ao software AutoCAD 13 para desenho. Os dados assim obtidos permitiram a geração de plantas e de cálculo de distâncias horizontais e áreas nos três casos. Foi possível se plotar as três plantas em único desenho por se considerar para o primeiro ponto da poligonal obtida por levantamento topográfico convencional, o mesmo par de coordenadas que o obtido pelo receptor GNSS. Concluiu-se que valor de área obtida através do Google Earth ficando próximo do valor da referência, bem como seu entorno, não significa a precisão do polígono visto que comprometedoras diferenças em distâncias ora à maior ora à menor podem estar sendo compensadas, no entanto pode-se admitir esse procedimento para uso em planejamento rural generalizado. / Abstract: The purpose of this study was to compare horizontal distances and areas of a polygon considering homologous points obtained through conventional survey using a Nikon Nivo total station 322d, a GNSS receiver AshTech Pro Mark 200 for georeferenced survey and Google Earth image. The processing of the survey made by computer program Datageosis Office version accused high precision constituting as the reference. The data obtained from the GNSS receiver were post-processed by the GNSS software Solution and the data obtained by Google Earth was submitted to the 13 AutoCAD software for drawing. The data obtained enabled the maps generation and calculation horizontal distances and areas in all three cases. It was possible to plot the three maps in a unique design by considering for the first point of the polygon obtained by conventional surveying, the same pair of coordinates obtained by the GPS receiver. It was concluded that the area value obtained through Google Earth being next to the reference value, does not means the polygon precision due to the fact of differences in distances sometimes the largest and sometimes the smallest being compensated, however it is possible to admit this procedure for use in general rural planning / Mestre

Thesis (M.Sc. (Soil Science)) --University of Limpopo, 2001 / Refer to document

Bibliography: p. 231-232. / xvi, 232 p. : ill. ; 30 cm. / Title page, contents and abstract only. The complete thesis in print form is available from the University Library. / The aim of this thesis is detecting and tracking objects at large ranges, when no target features are visible, with imaging type sensors. A system which estimates the optical flow of the scene in a parallel architecture is developed. The architecture is similar to that of an artifical neural network. / Thesis (Ph.D.)--University of Adelaide, Dept. of Electrical and Electronic Engineering, 1994

Extraction of buildings from remote sensing sources is an important GIS application and has been the subject of extensive research over the last three decades. An accurate building inventory is required for applications such as GIS database maintenance and revision; impervious surfaces mapping; storm water management; hazard mitigation and risk assessment. Despite all the progress within the fields of photogrammetry and image processing, the problem of automated feature extraction is still unresolved. A methodology for automatic building extraction that integrates remote sensing sources and GIS data was proposed. The methodology consists of a series of image processing and spatial analysis techniques. It incorporates initial simplification procedure and multiple feature analysis components. The extraction process was implemented and tested on three distinct types of buildings including commercial, residential and high-rise. Aerial imagery and GIS data from Shelby County, Tennessee were identified for the testing and validation of the results. The contribution of each component to the overall methodology was quantitatively evaluated as relates to each type of building. The automatic process was compared to manual building extraction and provided means to alleviate the manual procedure effort. A separate module was implemented to identify the 2D shape of a building. Indices for two specific shapes were developed based on the moment theory. The indices were tested and evaluated on multiple feature segments and proved to be successful. The research identifies the successful building extraction scenarios as well as the challenges, difficulties and drawbacks of the process. Recommendations are provided based on the testing and evaluation for future extraction projects.

Thesis (Ph. D.)--Mississippi State University. Department of Electrical and Computer Engineering. / Title from title screen. Includes bibliographical references.

Meteorological satellite images at visible and infra-red wavelengths are an invaluable source of information on cloud systems because of their extensive coverage of the whole of the Earth's surface, providing data in areas that are only sparsely monitored, if at all, by other means. Although this information has been used subjectively by forecasters for many years, the lack of automatic, quantitative analysis techniques largely prevents its assimilation into numerical weather prediction (NWP) models, which are the basis of all modern weather forecasting. This thesis investigates the use of neural network techniques for the analysis of the images in order to make fuller use of the available data. The recognition of a particular type of cloud is dependent on the determination of a set of features from the satellite image spectral bands that will give discriminating information. This feature extraction and selection process is dealt with in detail, and a feature selection process based on the radial basis function (RBF) neural network is presented. The high-dimensional feature space is visualized on a two-dimensional plane by the use of three techniques: the Kohonen map, the Sammon map, and a recently-developed technique known as the Generative Topographic Mapping (GTM). Classification results using a multi-layer perceptron (MLP) and an RBF neural network are presented. The results of independently classifying each pixel in an image are compared with a method that makes use of contextual information, the Markov Random Field (MRF) model. The limitations of the pixel-based approach are highlighted, and a region-based approach is presented that enables the definition of large-scale regions of uniform cloud type. Two segmentation methods are used, the active contour (or snake) model, and the more recentlydeveloped level set technique. The latter method was found to provide many benefits over the former. The region-based approach will facilitate the assimilation of cloud system information into NWP models in the future.