Acurácia de classes altimétricas geradas a partir de produtos de sensores de radar
Carregando...
Data
2016-07-29
Autores
Rodrigues, Bruno Timóteo [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Os novos sensores orbitais e a aquisição contínua de imagens da superfície terrestre têm oferecido um volume e variedade cada vez maior de dados a comunidade de sensoriamento remoto. Como exemplo desse novo segmento da modelagem ambiental, o projeto SRTM assim como o ASTER, obtendo como produtos Modelos Digitais de Terreno (MDT) de até 80% da área terrestre do planeta. Este trabalho teve como objetivo avaliar a qualidade e precisão de dados altimétricos obtidos em Modelos Digitais de Elevação (MDEs). Para tanto, foi elaborado um mapa altimétrico base (Verdade terrestre), o qual foi comparado, por analise do Indice Kappa, com três modelos digitais de terreno (MDT) obtidos dos projetos TOPODATA, SRTM 30 metros e ASTER. De acordo com a avaliação expressa pela Matriz de Confusão gerada a partir da Altimetría do TOPODATA cruzada a Verdade Terrestre, 19 das 26 classes de informações altimétricas apresentadas não tiveram os seus maiores valores expressos nos respectivos eixos principais das classes confrontadas com a verdade terrestre da área, constatando um bom grau de acuracidade nos dados classificados do cruzamento do TOPODATA com a Verdade Terrestre. Observou-se que as classes que apresentaram os maiores valores um nível abaixo do eixo principal da matriz de confusão se localizam exatamente na região que corresponde à área com declividade muito acentuada, se tratando da feição geomorfológica da Cuesta. Desta forma considera-se que a resolução espacial das imagens provenientes dos sensores de radar, neste caso o arquivo do TOPODADA, está diretamente ligada com a baixa acurácia em áreas de grandes declividades, pois quanto maior o tamanho do pixel, maior a dificuldade de capturar diferenças abruptas de declividade. No Modelo digital de elevação do sensor de Radar SRTM 30m, apenas 2 das 26 classes de informações altimétricas apresentadas não tiveram os seus maiores valores expressos nos respectivos eixos principais das classes confrontadas com a verdade terrestre da área. Dessa forma, pode-se verificar um bom grau de acuracidade nos dados classificados. Para os dados analisados do sensor ASTER confrontados com a Verdade Terrestre, assim como os dados do SRTM 30m, somente 2 das 26 classes de informações altimétricas apresentadas não tiveram os seus maiores valores expressos nos respectivos eixos principais das classes confrontadas com a verdade terrestre. Desta forma, também considera um bom grau de acuracidade do cruzamento do dados classificados dos sensores ASTER com a Verdade Terrestre. Por fim, observou-se que o sensor de radar SRTM 30 metros exibiu o melhor dos MDEs analisados, por apresentar resultados de índice Kappa, com 3,798% de acertos a mais em relação ao sensor ASTER e 6,43% a mais que o dados do TOPODATA.
The new orbital sensors and continuous acquisition of land surface images have offered a volume and increasing variety of data to remote sensing community. As an example of this new segment of the environmental modeling, SRTM project as well as ASTER, obtaining as products Digital Terrain Models (DTM) of up to 80% of the Earth's land area. To evaluate the quality and accuracy of altimetry data from Digital Elevation Models (DEMs). Thus, a altimetry basemap (terrestrial Truth) was prepared which was compared by analysis of Kappa Index, with three digital terrain models (DTM) obtained from TOPODATA projects, SRTM and ASTER 30 meters. According to the evaluation expressed by the confusion matrix generated from the altimetry of cross TOPODATA Truth Land, 19 of the 26 classes presented altimetric information did not have their greatest values expressed in the respective main axes of the classes faced with the ground truth area , finding a good degree of accuracy in the data classified as TOPODATA the intersection with Road Truth. It was observed that the classes that showed the highest values one level below the main confusion matrix axis is located exactly in the region that corresponds to the area with very steep slopes, in the case of geomorphological feature of Cuesta. Thus it is considered that the spatial resolution from images of the radar sensors, in this case TOPODADA file is directly connected with the low accuracy in areas of great slope, because the larger the pixel size, the more difficult to capture sharp differences in steepness. In the digital elevation model of the radar sensor SRTM 30m, only 2 of the 26 classes of altimetric information submitted have not had their highest values expressed in the respective main axes of the classes faced with the ground truth of the area. Thus, it can be seen a good degree of accuracy in the data sorted. For the analyzed data ASTER faced with the Earth Truth, as well as data from SRTM 30m, only 2 of the 26 classes of altimetric information submitted have not had their highest values expressed in the respective main axes of the classes faced with the ground truth. Thus, also considers a good degree of accuracy of the data crossing the classified ASTER sensors with the Earth Truth. Finally, there was the SRTM 30 meters radar sensor exhibited the best of MOUs analyzed by presenting the results of Kappa index, 3.798% of the more hits against the ASTER and 6.43% more than the data the TOPODATA.
The new orbital sensors and continuous acquisition of land surface images have offered a volume and increasing variety of data to remote sensing community. As an example of this new segment of the environmental modeling, SRTM project as well as ASTER, obtaining as products Digital Terrain Models (DTM) of up to 80% of the Earth's land area. To evaluate the quality and accuracy of altimetry data from Digital Elevation Models (DEMs). Thus, a altimetry basemap (terrestrial Truth) was prepared which was compared by analysis of Kappa Index, with three digital terrain models (DTM) obtained from TOPODATA projects, SRTM and ASTER 30 meters. According to the evaluation expressed by the confusion matrix generated from the altimetry of cross TOPODATA Truth Land, 19 of the 26 classes presented altimetric information did not have their greatest values expressed in the respective main axes of the classes faced with the ground truth area , finding a good degree of accuracy in the data classified as TOPODATA the intersection with Road Truth. It was observed that the classes that showed the highest values one level below the main confusion matrix axis is located exactly in the region that corresponds to the area with very steep slopes, in the case of geomorphological feature of Cuesta. Thus it is considered that the spatial resolution from images of the radar sensors, in this case TOPODADA file is directly connected with the low accuracy in areas of great slope, because the larger the pixel size, the more difficult to capture sharp differences in steepness. In the digital elevation model of the radar sensor SRTM 30m, only 2 of the 26 classes of altimetric information submitted have not had their highest values expressed in the respective main axes of the classes faced with the ground truth of the area. Thus, it can be seen a good degree of accuracy in the data sorted. For the analyzed data ASTER faced with the Earth Truth, as well as data from SRTM 30m, only 2 of the 26 classes of altimetric information submitted have not had their highest values expressed in the respective main axes of the classes faced with the ground truth. Thus, also considers a good degree of accuracy of the data crossing the classified ASTER sensors with the Earth Truth. Finally, there was the SRTM 30 meters radar sensor exhibited the best of MOUs analyzed by presenting the results of Kappa index, 3.798% of the more hits against the ASTER and 6.43% more than the data the TOPODATA.
Descrição
Palavras-chave
Modelo Digital de Elevação, Sensoriamento Remoto, Geoprocessamento, Matriz de confusão, Índice Kappa, Digital Elevation Model, Remote Sensing, Geoprocessing, Matrix Confusion, Kappa Index