Desenvolvimento e avaliação de uma metodologia para detecção de mudanças na cobertura vegetal do semi-árido. / Development and evaluation of a methodology for detection of changes in the vegetal cover of the semi-arid one.

AUTOR(ES)
DATA DE PUBLICAÇÃO

2004

RESUMO

In this work a methodology for the change detection in vegetation cover of semi-arid region was developed, using multi-temporal images of TMLandsat, in the red spectral band. In order to implement operationally this methodology two digital techniques were developed. These techniques do not need to make use of images with atmospheric correction for the generation of change maps. We used the simple concept of "radiometric axis of no change", which allows to control the digital algorithm of change detection, using field information for the labeled samples. These techniques were called Radiometric Rotation Controlled by Axis of No Change (RCEN) and Radiometric Rotation Controlled by Accuracy Index (RCIA). These techniques extract samples of radiometric values from multi-temporal images, establishing parameters that allow rotating the radiometric values of the images, to generate the change detection image. Afterwards a slicing is made and labeled from the detected image, what, on the basis of field information, allows the generation of a final map of changes. In order to validate the capacity of those techniques it was made a comparison results from the Simple difference with the Principal Components Analysis (PCA). The selected areas for this study are located in the central region of the Brazilian semi-arid tropic (Quixaba-PE) (these different periods were analysed between 1984 and 2001) and in the center-west region of the Argentinean semi-arid (Valle Fértil - San Juan), in the timeframe of 1987- 2001. Four classes of changes and one of no-change were defined in this multitemporal analysis, and the thematic accuracy was evaluated, considering the matrix of confusion and Kappa analysis. Consistent results were obtained with the RCEN technique (total accuracy > 87% and K index > 0.70 in average), maintaining significant performance in all the investigated time intervals. The methodology developed in this thesis reduces the complexity of the digital procedure of images, and that it besides that it allow to the spatial representation on dynamics of use and land cover, in the semi-arid region. The cartography of these changes is extremely necessary to the planning of land use and conservation of the natural resources of the so-called region "South American semi-arid diagonal".

ASSUNTO(S)

quixaba (pe) sensoriamento remoto novas técnicas change detection new techniques monitoriamento ambiental detecção de mudanças remote sensing arid lands terras aridas environmental moniroring san juan (argentina) san juan (argentina) quixaba (pe)

ACESSO AO ARTIGO

Documentos Relacionados