The need to quantify distance between two groups of objects is prevalent throughout the signal processing world. The difference of group means computed using the Euclidean, or L2 distance, is one of the predominant distance measures used to compare feature vectors and groups of vectors, but many problems arise with it when high data dimensionality is present. Maximum mean discrepancy (MMD) is a recent unsupervised kernel-based pattern recognition method which may improve differentiation between two distinct populations over many commonly used methods such as the difference of means, when paired with the proper feature representations and kernels. MMD-based distance computation combines many powerful concepts from the machine learning literature, such as data distribution-leveraging similarity measures and kernel methods for machine learning. Due to this heritage, we posit that dissimilarity-based classification and changepoint detection using MMD can lead to enhanced separation between different populations. To test this hypothesis, we conduct studies comparing MMD and the difference of means in two subareas of image analysis and understanding: first, to detect scene changes in video in an unsupervised manner, and secondly, in the biomedical imaging field, using clinical ultrasound to assess tumor response to treatment. We leverage effective computer vision data descriptors, such as the bag-of-visual-words and sparse combinations of SIFT descriptors, and choose from an assessment of several similarity kernels (e.g. Histogram Intersection, Radial Basis Function) in order to engineer useful systems using MMD. Promising improvements over the difference of means, measured primarily using precision/recall for scene change detection, and k-nearest neighbour classification accuracy for tumor response assessment, are obtained in both applications.

Technical advances in network communication systems (e.g. radio access networks) combined with evolving concepts based on virtualization (e.g. clouds), require new management algorithms in order to handle the increasing complexity in the network behavior and variability in the network environment. Current network management operations are primarily centralized and deterministic, and are carried out via automated scripts and manual interventions, which work for mid-sized and fairly static networks. The next generation of communication networks and systems will be of significantly larger size and complexity, and will require scalable and autonomous management algorithms in order to meet operational requirements on reliability, failure resilience, and resource-efficiency. A promising approach to address these challenges includes the development of probabilistic management algorithms, following three main design goals. The first goal relates to all aspects of scalability, ranging from efficient usage of network resources to computational efficiency. The second goal relates to adaptability in maintaining the models up-to-date for the purpose of accurately reflecting the network state. The third goal relates to reliability in the algorithm performance in the sense of improved performance predictability and simplified algorithm control. This thesis is about probabilistic approaches to fault management that follow the concepts of probabilistic network management (PNM). An overview of existing network management algorithms and methods in relation to PNM is provided. The concepts of PNM and the implications of employing PNM-algorithms are presented and discussed. Moreover, some of the practical differences of using a probabilistic fault detection algorithm compared to a deterministic method are investigated. Further, six probabilistic fault management algorithms that implement different aspects of PNM are presented. The algorithms are highly decentralized, adaptive and autonomous, and cover several problem areas, such as probabilistic fault detection and controllable detection performance; distributed and decentralized change detection in modeled link metrics; root-cause analysis in virtual overlays; event-correlation and pattern mining in data logs; and, probabilistic failure diagnosis. The probabilistic models (for a large part based on Bayesian parameter estimation) are memory-efficient and can be used and re-used for multiple purposes, such as performance monitoring, detection, and self-adjustment of the algorithm behavior. / <p>QC 20140509</p>

Thesis (MSc (Geography and Environmental Studies))--University of Stellenbosch, 2005. / Land degradation in Lesotho is undermining the finite resource on which people depend for survival. Use of satellite imagery has been recommended for monitoring land degradation because remotely sensed data enable monitoring of large areas at more frequent intervals than intensive ground based research. Various techniques have been developed for land cover change detection. In the present study, vegetation changes were identified by image differencing, which involved finding the difference between the earlier date NDVI image and the later date image. NDVI images are among products that are generated from the NOAA AVHRR sensor to provide information about the quantity of biomass on the earth’s surface. The resulting NDVI change data showed land areas that had experienced vegetation loss, which were identified as potentially degraded. The change data were combined with other data sets to determine how potentially degraded areas were influenced by different environmental variables and population pressure. These data sets included land cover, ecological zones, elevation, soil and human and livestock populations. By integrating NDVI data with ancillary data, land degradation was attributed to both demographic pressure and biophysical factors. Widespread degradation was detected on the arable parts of the Lowlands where cultivation was intensive and human settlements were extensive. Signs of grassland depletion and forest decline were also evident and were attributed to population expansion, overgrazing and indiscriminate cutting of trees and shrubs for firewood. Extensive biomass decline was also associated more with soils in the lowlands derived from sedimentary rocks than soils of basalt origin that occur mostly in the highlands. Significant degradation was evident on gentle slopes where land uses such as cultivation and expansion of settlements were identified as the main causes of the degradation. There was evidence of greater vegetation depletion on north and east-facing slopes than on other slopes. The depletion was attributed to the fragility of ecosystems resulting from intense solar radiation. The study demonstrated that NOAA AVHRR NDVI images could be used effectively for detecting land cover changes in Lesotho. However, future research could focus on obtaining and using high resolution data for detailed analysis of factors driving land degradation.

A detecção de mudanças no uso e na cobertura do solo pode ser considerada a função do sensoriamento remoto que agrega uma dimensão temporal à análise das informações contidas nas imagens. Ao confrontá-las duas a duas, para identificar, localizar e qualificar as transformações que ocorreram na cobertura e no uso do solo em determinados espaço e tempo, através das respostas espectrais registradas nos pares de pixels homólogos quando aplicados limiares que permitam distinguir entre a mudança e a não mudança. A análise ponto a ponto, ou instante a instante permite inferir sobre a quantidade e qualidade das mudanças detectadas em uma região durante um determinado período. Já a análise comparativa entre dois ou mais destes resultados, utilizando uma série temporal de imagens, permite inferir acerca da dinâmica das transformações em vários outros aspectos; como topologia, intensidade, tipo de mudança (substituição ou conversão) e taxa de mudança. São várias as técnicas disponíveis para a detecção de mudanças no uso e cobertura do solo a partir de imagens digitais, obtidas por sensores orbitais, dois grandes grupos podem ser propostos: técnicas de pré-classificação e de pós-classificação, diferindo fundamentalmente sobre quais produtos são aplicados os limiares da detecção das mudanças, se sobre produtos temáticos da classificação de imagens, ou se sobre imagens diretamente. Este trabalho utiliza técnicas de detecção baseadas em subtração de imagens de ambos os grupos, especificamente, a pós-classificação por máxima verossimilhança e da pré-classificação, por Índice de Vegetação por Diferença Normalizada (NDVI) e por Transformada Kauth-Thomas (KT), nesse caso o componente de verdor. Visa avaliar a sensibilidade e adequação destas técnicas para a análise das transformações ocorridas no uso e da cobertura do solo durante os dois períodos de comparação: 1988 a 2001 e 2001 a 2006 e no conjunto dos 18 anos, na captação das tendências das transformações deste ambiente da Campanha Sul do Estado do Rio Grande do Sul, que é majoritariamente rural, muito dinâmico e bastante heterogêneo. Considerando que durante o período abrangido neste trabalho, grandes fazendas tradicionais de pecuária, em um primeiro momento, foram convertidas para agricultura familiar através da criação intensiva de assentamentos da reforma agrária, ocorrida entre a metade dos anos 80 e a metade dos anos 90, mais recentemente, substituídas por plantios florestais da indústria do papel. Os resultados permitiram confrontar os tratamentos e verificar as acurácias das detecções e identificar as principais dificuldades, em especial, o efeito da fenologia nas diversas fases em que se apresentam nas substituições florestais de ciclos longos. A dificuldade da técnica KT, em lidar com plantios semi-perenes e perenes, a impossibilidade de se considerar áreas cobertas em algum momento por nuvens. Por fim, corrobora com a inviabilidade do estabelecimento a priori da melhor técnica, ou mesmo, daquela mais acurada, sem que sejam considerados plenamente os objetivos, a escala, a natureza do ambiente analisado e as classes de mudança estabelecidas para o trabalho, além da qualidade das imagens disponíveis. / The detection of alterations in land use and cover can be considered as being an operation in Remote Sensing which adds a time dimension to the analysis of information in images. This is done when images are compared, by groups of two, at certain space and time looking for spectral responses stored in pairs of homologous pixels, through the application of thresholds which lead to the differentiation between change and non-change. A point-to-point, or instant-to-instant analysis, permits to infer on the amount and quality of alterations detected in a region, during a certain period. The comparative analysis between two or more of these results, via a time series of images, informs on the dynamics of transformations in other aspects, as topology, intensity, kind of change (substitution or conversion), and change rate. Several techniques are available to detect alterations in land use and cover, from digital images collected by orbital sensors. Two larger groups can be highlighted: preclassification techniques, and post-classification techniques. They differ basically on over which products the thresholds defining changes are applied, these products being either thematic ones for image classification, or the image itself. This work uses detection techniques based on image subtraction of both groups. Pre-classification uses the Normalized Difference Vegetation Index (NDVI) and the Kauth-Thomas Transform (KT), the green index in this last case. Post-classification uses the Maximum Likelihood. The objective is to estimate the sensitivity and adequacy of these techniques for the detection and analysis of changes in land use and cover during two comparison periods: the first one is from 1988 to 2001; the second, from 2001 to 2006. Besides, the whole 18-years period is studied to detect tendencies of the transformation of the region. The study area is at the Campanha Sul region, at Rio Grande do Sul State Brazil. It is largely rural, heterogeneous and dynamic, since during the period covered (1988-2006) large estates where either converted into smaller properties, family-managed; though an intensive policy of agrarian reform (from the mid-eighties to the mid-nineties), or, more recently, by industrial-style cultures of forests to serve the paper industry. The results allowed comparing the different treatments and to verify the accuracy of detections. The main difficulties were the phonological cycles, the various phases of long-cycle artificial forests, the limitations of the KT technique to handle semi-perennial cultures, and cloud-covered areas. It was not possible to clearly define the better or more accurate technique; this definition depends of specific objectives, of the scale and nature of the study region, and of the classes of change being analyzed, besides of the quality of available images.

As Unidades de Conservação (UC) são espaços territoriais com características naturais relevantes, que têm a função de assegurar a representatividade de amostras significativas e ecologicamente viáveis das diferentes populações, habitats e ecossistemas. A Legislação Brasileira instituiu no ano de 2000 o Sistema Nacional de Unidades de Conservação da Natureza (SNUC). Dentre os diversos ambientes encontrados em território nacional, o Pampa tem uma representatividade de apenas 0,4% de sua área protegida, conforme consta no SNUC. O Refúgio da Vida Silvestre Banhado dos Pachecos (RVSBP) é uma UC de proteção integral estadual, localizada no Rio Grande do Sul e no Bioma Pampa, com uma área de 2.560 ha e representa cerca de 3,5% de todas as UC’s de proteção integral desse bioma. O RVSBP criada no ano de 2002, e ainda não possui plano de manejo, bem como carece de maiores investimentos e atenção. O uso de imagens de satélites como subsídio aos estudos ambientais já está consolidado e a interpretação destas imagens, a partir de diversos métodos, para classificar o uso e cobertura da terra, tem se tornado uma constante, munindo os pesquisadores de informações dos diversos processos que possam estar ocorrendo em uma determinada área de estudo, inclusive monitorando as mudanças ao longo do tempo. Os objetivos desta dissertação são os de verificar as mudanças no uso e cobertura do solo ocorridas entre 2001 e 2017 no RVSBP e em seu entorno direto de 10km, baseando-se na análise de imagens de satélite. Para tanto serão mapeadas as classes de uso e cobertura do solo, a partir de imagens dos satélites LANDSAT 5 – Sensor TM, LANDSAT 7 – Sensor ETM+ e LANDSAT 8 – Sensor OLI, para os anos de 2001, 2009 e 2017. O método de detecção das mudanças no uso e cobertura do solo aplicada foi a técnica de comparação pós-classificação para uma melhor compreensão das interações entre os fenômenos naturais e as atividades humanas. Essa técnica foi aplicada para os períodos de 2001 a 2009, de 2009 a 2017 e de 2001 a 2017. Para o período de 2001 a 2009 as mudanças ocorreram em 17,5% da área de estudo e em 19,9% do RVSBP. Para o período de 2009 a 2017 as mudanças ocorreram em 22,8% da área de estudo e em 23,9% do RVSBP. Para o período de 2001 a 2017 as mudanças ocorreram em 24% da área de estudo e em 32% do RVSBP. Dentre esses 32% a classe que apresentou os maiores acréscimos de área foram as classes de Agricultura – Arroz e de Associação de Sítio e produtores rurais, que respectivamente compreendem áreas de 410 hectares e de 135 hectares. As classes que foram mais impactadas com perda de área foram as classes Banhado e Vegetação Arbórea, que respectivamente compreendem áreas de 435 hectares e de 173 hectares. A análise de detecção de mudanças se mostrou efetiva como uma forma de monitoramento sistemático do uso e cobertura do solo do RVSBP e entorno, trazendo elementos importantes para a gestão da UC. / Conservation Units (UC) are territorial spaces with relevant natural characteristics, which have a role of ensuring the representativeness of significant and ecologically viable samples of different populations, habitats and ecosystems. Brazilian legistlation established in 2000 the National System of Nature Conservation Units (SNUC). Among the several environments found in the national territory, the Pampa has a representation of only 0,4% of its own protected area, according to SNUC. The Wildlife Refuge Banhado dos Pachecos (RVSBP) is a state UC of integral protection, located in Rio Grande do Sul and Bioma Pampa, with 2.560 ha and comprises about 3,5% of all integral protection UC of this biome. RVSBP was created in 2002, still does not have a management plan, and lacks greater investments and attention. The use of satellite images to suppott environmental studies is already consolidate and the interpretation of these images, using different methods, to classify land and use cover, has become a constant, providing researchers with information on the various processes that may be occurring in a particular study area, including monitoring changes over time. The objective of this dissertation is to verify the changes in the land use and cover occurred between 2001 and 2017 in RVSBP and in its surrounds of 10km, based on the analysis of satellite images. Therefore, the land and use coverage classes were mapped using images from the LANDSAT 5 - Sensor TM, LANDSAT 7 - ETM + and LANDSAT 8 - OLI Sensor, for the years 2001, 2009 and 2017. The method of detecting changes in land use and cover was the post-classification comparison technique for a better understanding of the interactions between natural phenomena and human activities. This technique was applied for the periods from 2001 to 2009, from 2009 to 2017, and from 2001 to 2017. For the period 2001 to 2009 the changes occurred in 17,5% of the whole study area and in 19,9% of RVSBP. For the period from 2009 to 2017, changes occurred in 22,8% of the whole study area and 23,9% of RVSBP. For the period from 2001 to 2017, changes occurred in 24% of the whole study area and 32% of RVSBP. Among these 32%, the class with the greatest increases in area were Agriculture – Rice crops and Site Association of Rural Producers, which respectively comprises areas of 410 hectares and 135 hectares. The classes that were most impacted with loss of area were the class Weands and Arboreal Vegetation, which respectively comprise areas of 435 and 173 hectares. The change detection analysis was effective as a way of systematically monitoring the land use and coverage of RVSBP and surroundings, bringing important elements to the management of the UC.

A detecção de mudanças no uso e na cobertura do solo pode ser considerada a função do sensoriamento remoto que agrega uma dimensão temporal à análise das informações contidas nas imagens. Ao confrontá-las duas a duas, para identificar, localizar e qualificar as transformações que ocorreram na cobertura e no uso do solo em determinados espaço e tempo, através das respostas espectrais registradas nos pares de pixels homólogos quando aplicados limiares que permitam distinguir entre a mudança e a não mudança. A análise ponto a ponto, ou instante a instante permite inferir sobre a quantidade e qualidade das mudanças detectadas em uma região durante um determinado período. Já a análise comparativa entre dois ou mais destes resultados, utilizando uma série temporal de imagens, permite inferir acerca da dinâmica das transformações em vários outros aspectos; como topologia, intensidade, tipo de mudança (substituição ou conversão) e taxa de mudança. São várias as técnicas disponíveis para a detecção de mudanças no uso e cobertura do solo a partir de imagens digitais, obtidas por sensores orbitais, dois grandes grupos podem ser propostos: técnicas de pré-classificação e de pós-classificação, diferindo fundamentalmente sobre quais produtos são aplicados os limiares da detecção das mudanças, se sobre produtos temáticos da classificação de imagens, ou se sobre imagens diretamente. Este trabalho utiliza técnicas de detecção baseadas em subtração de imagens de ambos os grupos, especificamente, a pós-classificação por máxima verossimilhança e da pré-classificação, por Índice de Vegetação por Diferença Normalizada (NDVI) e por Transformada Kauth-Thomas (KT), nesse caso o componente de verdor. Visa avaliar a sensibilidade e adequação destas técnicas para a análise das transformações ocorridas no uso e da cobertura do solo durante os dois períodos de comparação: 1988 a 2001 e 2001 a 2006 e no conjunto dos 18 anos, na captação das tendências das transformações deste ambiente da Campanha Sul do Estado do Rio Grande do Sul, que é majoritariamente rural, muito dinâmico e bastante heterogêneo. Considerando que durante o período abrangido neste trabalho, grandes fazendas tradicionais de pecuária, em um primeiro momento, foram convertidas para agricultura familiar através da criação intensiva de assentamentos da reforma agrária, ocorrida entre a metade dos anos 80 e a metade dos anos 90, mais recentemente, substituídas por plantios florestais da indústria do papel. Os resultados permitiram confrontar os tratamentos e verificar as acurácias das detecções e identificar as principais dificuldades, em especial, o efeito da fenologia nas diversas fases em que se apresentam nas substituições florestais de ciclos longos. A dificuldade da técnica KT, em lidar com plantios semi-perenes e perenes, a impossibilidade de se considerar áreas cobertas em algum momento por nuvens. Por fim, corrobora com a inviabilidade do estabelecimento a priori da melhor técnica, ou mesmo, daquela mais acurada, sem que sejam considerados plenamente os objetivos, a escala, a natureza do ambiente analisado e as classes de mudança estabelecidas para o trabalho, além da qualidade das imagens disponíveis. / The detection of alterations in land use and cover can be considered as being an operation in Remote Sensing which adds a time dimension to the analysis of information in images. This is done when images are compared, by groups of two, at certain space and time looking for spectral responses stored in pairs of homologous pixels, through the application of thresholds which lead to the differentiation between change and non-change. A point-to-point, or instant-to-instant analysis, permits to infer on the amount and quality of alterations detected in a region, during a certain period. The comparative analysis between two or more of these results, via a time series of images, informs on the dynamics of transformations in other aspects, as topology, intensity, kind of change (substitution or conversion), and change rate. Several techniques are available to detect alterations in land use and cover, from digital images collected by orbital sensors. Two larger groups can be highlighted: preclassification techniques, and post-classification techniques. They differ basically on over which products the thresholds defining changes are applied, these products being either thematic ones for image classification, or the image itself. This work uses detection techniques based on image subtraction of both groups. Pre-classification uses the Normalized Difference Vegetation Index (NDVI) and the Kauth-Thomas Transform (KT), the green index in this last case. Post-classification uses the Maximum Likelihood. The objective is to estimate the sensitivity and adequacy of these techniques for the detection and analysis of changes in land use and cover during two comparison periods: the first one is from 1988 to 2001; the second, from 2001 to 2006. Besides, the whole 18-years period is studied to detect tendencies of the transformation of the region. The study area is at the Campanha Sul region, at Rio Grande do Sul State Brazil. It is largely rural, heterogeneous and dynamic, since during the period covered (1988-2006) large estates where either converted into smaller properties, family-managed; though an intensive policy of agrarian reform (from the mid-eighties to the mid-nineties), or, more recently, by industrial-style cultures of forests to serve the paper industry. The results allowed comparing the different treatments and to verify the accuracy of detections. The main difficulties were the phonological cycles, the various phases of long-cycle artificial forests, the limitations of the KT technique to handle semi-perennial cultures, and cloud-covered areas. It was not possible to clearly define the better or more accurate technique; this definition depends of specific objectives, of the scale and nature of the study region, and of the classes of change being analyzed, besides of the quality of available images.

As Unidades de Conservação (UC) são espaços territoriais com características naturais relevantes, que têm a função de assegurar a representatividade de amostras significativas e ecologicamente viáveis das diferentes populações, habitats e ecossistemas. A Legislação Brasileira instituiu no ano de 2000 o Sistema Nacional de Unidades de Conservação da Natureza (SNUC). Dentre os diversos ambientes encontrados em território nacional, o Pampa tem uma representatividade de apenas 0,4% de sua área protegida, conforme consta no SNUC. O Refúgio da Vida Silvestre Banhado dos Pachecos (RVSBP) é uma UC de proteção integral estadual, localizada no Rio Grande do Sul e no Bioma Pampa, com uma área de 2.560 ha e representa cerca de 3,5% de todas as UC’s de proteção integral desse bioma. O RVSBP criada no ano de 2002, e ainda não possui plano de manejo, bem como carece de maiores investimentos e atenção. O uso de imagens de satélites como subsídio aos estudos ambientais já está consolidado e a interpretação destas imagens, a partir de diversos métodos, para classificar o uso e cobertura da terra, tem se tornado uma constante, munindo os pesquisadores de informações dos diversos processos que possam estar ocorrendo em uma determinada área de estudo, inclusive monitorando as mudanças ao longo do tempo. Os objetivos desta dissertação são os de verificar as mudanças no uso e cobertura do solo ocorridas entre 2001 e 2017 no RVSBP e em seu entorno direto de 10km, baseando-se na análise de imagens de satélite. Para tanto serão mapeadas as classes de uso e cobertura do solo, a partir de imagens dos satélites LANDSAT 5 – Sensor TM, LANDSAT 7 – Sensor ETM+ e LANDSAT 8 – Sensor OLI, para os anos de 2001, 2009 e 2017. O método de detecção das mudanças no uso e cobertura do solo aplicada foi a técnica de comparação pós-classificação para uma melhor compreensão das interações entre os fenômenos naturais e as atividades humanas. Essa técnica foi aplicada para os períodos de 2001 a 2009, de 2009 a 2017 e de 2001 a 2017. Para o período de 2001 a 2009 as mudanças ocorreram em 17,5% da área de estudo e em 19,9% do RVSBP. Para o período de 2009 a 2017 as mudanças ocorreram em 22,8% da área de estudo e em 23,9% do RVSBP. Para o período de 2001 a 2017 as mudanças ocorreram em 24% da área de estudo e em 32% do RVSBP. Dentre esses 32% a classe que apresentou os maiores acréscimos de área foram as classes de Agricultura – Arroz e de Associação de Sítio e produtores rurais, que respectivamente compreendem áreas de 410 hectares e de 135 hectares. As classes que foram mais impactadas com perda de área foram as classes Banhado e Vegetação Arbórea, que respectivamente compreendem áreas de 435 hectares e de 173 hectares. A análise de detecção de mudanças se mostrou efetiva como uma forma de monitoramento sistemático do uso e cobertura do solo do RVSBP e entorno, trazendo elementos importantes para a gestão da UC. / Conservation Units (UC) are territorial spaces with relevant natural characteristics, which have a role of ensuring the representativeness of significant and ecologically viable samples of different populations, habitats and ecosystems. Brazilian legistlation established in 2000 the National System of Nature Conservation Units (SNUC). Among the several environments found in the national territory, the Pampa has a representation of only 0,4% of its own protected area, according to SNUC. The Wildlife Refuge Banhado dos Pachecos (RVSBP) is a state UC of integral protection, located in Rio Grande do Sul and Bioma Pampa, with 2.560 ha and comprises about 3,5% of all integral protection UC of this biome. RVSBP was created in 2002, still does not have a management plan, and lacks greater investments and attention. The use of satellite images to suppott environmental studies is already consolidate and the interpretation of these images, using different methods, to classify land and use cover, has become a constant, providing researchers with information on the various processes that may be occurring in a particular study area, including monitoring changes over time. The objective of this dissertation is to verify the changes in the land use and cover occurred between 2001 and 2017 in RVSBP and in its surrounds of 10km, based on the analysis of satellite images. Therefore, the land and use coverage classes were mapped using images from the LANDSAT 5 - Sensor TM, LANDSAT 7 - ETM + and LANDSAT 8 - OLI Sensor, for the years 2001, 2009 and 2017. The method of detecting changes in land use and cover was the post-classification comparison technique for a better understanding of the interactions between natural phenomena and human activities. This technique was applied for the periods from 2001 to 2009, from 2009 to 2017, and from 2001 to 2017. For the period 2001 to 2009 the changes occurred in 17,5% of the whole study area and in 19,9% of RVSBP. For the period from 2009 to 2017, changes occurred in 22,8% of the whole study area and 23,9% of RVSBP. For the period from 2001 to 2017, changes occurred in 24% of the whole study area and 32% of RVSBP. Among these 32%, the class with the greatest increases in area were Agriculture – Rice crops and Site Association of Rural Producers, which respectively comprises areas of 410 hectares and 135 hectares. The classes that were most impacted with loss of area were the class Weands and Arboreal Vegetation, which respectively comprise areas of 435 and 173 hectares. The change detection analysis was effective as a way of systematically monitoring the land use and coverage of RVSBP and surroundings, bringing important elements to the management of the UC.

A detecção de mudanças no uso e na cobertura do solo pode ser considerada a função do sensoriamento remoto que agrega uma dimensão temporal à análise das informações contidas nas imagens. Ao confrontá-las duas a duas, para identificar, localizar e qualificar as transformações que ocorreram na cobertura e no uso do solo em determinados espaço e tempo, através das respostas espectrais registradas nos pares de pixels homólogos quando aplicados limiares que permitam distinguir entre a mudança e a não mudança. A análise ponto a ponto, ou instante a instante permite inferir sobre a quantidade e qualidade das mudanças detectadas em uma região durante um determinado período. Já a análise comparativa entre dois ou mais destes resultados, utilizando uma série temporal de imagens, permite inferir acerca da dinâmica das transformações em vários outros aspectos; como topologia, intensidade, tipo de mudança (substituição ou conversão) e taxa de mudança. São várias as técnicas disponíveis para a detecção de mudanças no uso e cobertura do solo a partir de imagens digitais, obtidas por sensores orbitais, dois grandes grupos podem ser propostos: técnicas de pré-classificação e de pós-classificação, diferindo fundamentalmente sobre quais produtos são aplicados os limiares da detecção das mudanças, se sobre produtos temáticos da classificação de imagens, ou se sobre imagens diretamente. Este trabalho utiliza técnicas de detecção baseadas em subtração de imagens de ambos os grupos, especificamente, a pós-classificação por máxima verossimilhança e da pré-classificação, por Índice de Vegetação por Diferença Normalizada (NDVI) e por Transformada Kauth-Thomas (KT), nesse caso o componente de verdor. Visa avaliar a sensibilidade e adequação destas técnicas para a análise das transformações ocorridas no uso e da cobertura do solo durante os dois períodos de comparação: 1988 a 2001 e 2001 a 2006 e no conjunto dos 18 anos, na captação das tendências das transformações deste ambiente da Campanha Sul do Estado do Rio Grande do Sul, que é majoritariamente rural, muito dinâmico e bastante heterogêneo. Considerando que durante o período abrangido neste trabalho, grandes fazendas tradicionais de pecuária, em um primeiro momento, foram convertidas para agricultura familiar através da criação intensiva de assentamentos da reforma agrária, ocorrida entre a metade dos anos 80 e a metade dos anos 90, mais recentemente, substituídas por plantios florestais da indústria do papel. Os resultados permitiram confrontar os tratamentos e verificar as acurácias das detecções e identificar as principais dificuldades, em especial, o efeito da fenologia nas diversas fases em que se apresentam nas substituições florestais de ciclos longos. A dificuldade da técnica KT, em lidar com plantios semi-perenes e perenes, a impossibilidade de se considerar áreas cobertas em algum momento por nuvens. Por fim, corrobora com a inviabilidade do estabelecimento a priori da melhor técnica, ou mesmo, daquela mais acurada, sem que sejam considerados plenamente os objetivos, a escala, a natureza do ambiente analisado e as classes de mudança estabelecidas para o trabalho, além da qualidade das imagens disponíveis. / The detection of alterations in land use and cover can be considered as being an operation in Remote Sensing which adds a time dimension to the analysis of information in images. This is done when images are compared, by groups of two, at certain space and time looking for spectral responses stored in pairs of homologous pixels, through the application of thresholds which lead to the differentiation between change and non-change. A point-to-point, or instant-to-instant analysis, permits to infer on the amount and quality of alterations detected in a region, during a certain period. The comparative analysis between two or more of these results, via a time series of images, informs on the dynamics of transformations in other aspects, as topology, intensity, kind of change (substitution or conversion), and change rate. Several techniques are available to detect alterations in land use and cover, from digital images collected by orbital sensors. Two larger groups can be highlighted: preclassification techniques, and post-classification techniques. They differ basically on over which products the thresholds defining changes are applied, these products being either thematic ones for image classification, or the image itself. This work uses detection techniques based on image subtraction of both groups. Pre-classification uses the Normalized Difference Vegetation Index (NDVI) and the Kauth-Thomas Transform (KT), the green index in this last case. Post-classification uses the Maximum Likelihood. The objective is to estimate the sensitivity and adequacy of these techniques for the detection and analysis of changes in land use and cover during two comparison periods: the first one is from 1988 to 2001; the second, from 2001 to 2006. Besides, the whole 18-years period is studied to detect tendencies of the transformation of the region. The study area is at the Campanha Sul region, at Rio Grande do Sul State Brazil. It is largely rural, heterogeneous and dynamic, since during the period covered (1988-2006) large estates where either converted into smaller properties, family-managed; though an intensive policy of agrarian reform (from the mid-eighties to the mid-nineties), or, more recently, by industrial-style cultures of forests to serve the paper industry. The results allowed comparing the different treatments and to verify the accuracy of detections. The main difficulties were the phonological cycles, the various phases of long-cycle artificial forests, the limitations of the KT technique to handle semi-perennial cultures, and cloud-covered areas. It was not possible to clearly define the better or more accurate technique; this definition depends of specific objectives, of the scale and nature of the study region, and of the classes of change being analyzed, besides of the quality of available images.

Risks and uncertainties are unavoidable in agriculture in Sudan, due to its dependence on climatic factors and to the imperfect nature of the agricultural decisions and policies attributed to land cover and land use changes that occur. The current study was conducted in the Gash Agricultural Scheme (GAS) - Kassala State, as a semi-arid land in eastern Sudan. The scheme has been established to contribute to the rural development, to help stability of the nomadic population in eastern Sudan, particularly the local population around the Gash river areas, and to facilitate utilizing the river flood in growing cotton and other cash crops. In the last decade, the scheme production has declined, because of drought periods, which hit the region, sand invasion and the spread of invasive mesquite trees, in addition to administrative negligence. These have resulted also in poor agricultural productivity and the displacement of farmers away from the scheme area. Recently, the scheme is heavily disturbed by human intervention in many aspects. Consequently, resources of cultivated land have shrunk and declined during the period of the study, which in turn have led to dissatisfaction and increasing failure of satisfying increasing farmer’s income and demand for local consumption. Remote sensing applications and geospatial techniques have played a key role in studying different types of hazards whether they are natural or manmade. Multi-temporal satellite data combined with ancillary data were used to monitor, analyze and to assess land use and land cover (LULC) changes and the impact of land degradation on the scheme production, which provides the managers and decision makers with current and improved data for the purposes of proper administration of natural resources in the GAS. Information about patterns of LULC changes through time in the GAS is not only important for the management and planning, but also for a better understanding of human dimensions of environmental changes at regional scale. This study attempts to map and assess the impacts of LULC change and land degradation in GAS during a period of 38 years from 1972-2010. Dry season multi-temporal satellite imagery collected by different sensor systems was selected such as three cloud-free Landsat (MSS 1972, TM 1987 and ETM+ 1999) and ASTER (2010) satellite imagery. This imagery was geo-referenced and radiometrically and atmospherically calibrated using dark object subtraction (DOS). Two approaches of classification (object-oriented and pixel-based) were applied for classification and comparison of LULC. In addition, the study compares between the two approaches to determine which one is more compatible for classification of LULC of the GAS. The pixel-based approach performed slightly better than the object-oriented approach in the classification of LULC in the study area. Application of multi-temporal remote sensing data proved to be successful for the identification and mapping of LULC into five main classes as follows: woodland dominated by dense mesquite trees, grass and shrubs dominated by less dense mesquite trees, bare and cultivated land, stabilized fine sand and mobile sand. After image enhancement successful classification of imagery was achieved using pixel and object based approaches as well as subsequent change detection (image differencing and change matrix), supported by classification accuracy assessments and post-classification. Comparison of LULC changes shows that the land cover of GAS has changed dramatically during the investigated period. It has been discovered that more significant of LULC change processes occurred during the second studied period (1987 to 1999) than during the first period (1972-1987). In the second period nearly half of bare and cultivated lands was changed from 41372.74 ha (20.22 %) in 1987 to 28020.80 ha (13.60 %) in 1999, which was mainly due to the drought that hit the region during the mentioned period. However, the results revealed a drastic loss of bare and cultivated land, equivalent to more than 40% during the entire period (1972-2010). Throughout the whole period of study, drought and invasion of both mesquite trees and sand were responsible for the loss of more than 40% of the total productive lands. Change vector analysis (CVA) as a useful approach was applied for estimating change detection in both magnitude and direction of change. The promising approach of multivariate alteration detection (MAD) and subsequent maximum autocorrelation factor (MAD/MAF) transformation was used to support change detection via assessment of maximum correlation between the transformed variates and the specific original image bands related to specific land cover classes. However, both CVA and MAD/MAD strongly prove the fact that bare and cultivated land have dramatically changed and decreased continuously during the studied period. Both CVA and MAD/MAD demonstrate adequate potentials for monitoring, detecting, identifying and mapping the changes. Moreover, this research demonstrated that CVA and MAD/MAF are superior in providing qualitative details about the nature of all kinds of change. Vegetation indices (VI) such as normalized difference vegetation index (NDVI), soil-adjusted vegetation index (SAVI), modified adjusted vegetation index (MSAVI) and grain soil index (GSI) were applied to measure the quantitative characterization of temporal and spatial vegetation cover patterns and change. All indices remain very sensitive to structure variation of LULC. The results reveal that the NDVI is more effective for detecting the amount and status of the vegetation cover in the study area than SAVI, MSAVI and GSI. Therefore, it can be stated that NDVI can be used as a response variable to identify drought disturbance and land degradation in semi-arid land such as the GAS area. Results of detecting vegetation cover observed by using SAVI were found to be more reasonable than using MSAVI, although MSAVI reduces the background of bare soil better than SAVI. GSI proves high efficiency in determining the different types of surface soils, and producing a change map of top soil grain size, which is useful in assessment of land degradation in the study area. The linkage between socio-economic data and remotely sensed data was applied to determine the relationships between the different factors derived and to analyze the reasons for change in LULC and land degradation and its effects in the study area. The results indicate a strong relationship between LULC derived from remotely sensed data and the influencing socioeconomic variables. The results obtained from analyzing socioeconomic data confirm the findings of remote sensing data analysis, which assure that the decline and degradation of agricultural land is a result of further spread of mesquite trees and of increased invasion of sand during the study period. High livestock density and overgrazing, drought, invasion of sand, spread of invasive mesquite trees, overexploitation of land, improper management, and population growth were considered as the main direct factors responsible for degradation in the study area.

The main purpose of thesis is creation and evaluation of models for change detection of arable land to grassland by Hybrid-based Change Detection method, which combined approaches based on the Vegetation Indices, Image Differencing and Principal Component Analysis. Six locations with different seasonal configuration of images with high resolution and one locality covered by image with very high resolution were used. The areas were spread across the foothill areas of the Czech Republic. The selection of predictors and the most suitable model was supported by statistical calculation. Application selected models were carried out using a multi-temporal object classification and their accuracy were verified using reference data. The benefit of this thesis is finding generally applicable model useful to investigate the land cover change and evaluation of the potentially most appropriate seasonal configuration of images. Valuable is also methodology in this thesis which focus on selection of predictors and calculation the order of the most appropriate models, which is unique in the available literature. The thesis provides useful findings fitting to insufficiently explored issue of Change Detection arable land to grassland. Powered by TCPDF (www.tcpdf.org)