Close

@MastersThesis{Moreira:2000:InPoLa,
               author = "Moreira, Romero da Costa",
                title = "Influ{\^e}ncia do posicionamento e da largura de bandas de 
                         sensores remotos e dos efeitos atmosf{\'e}ricos na 
                         determina{\c{c}}{\~a}o de {\'{\i}}ndices de 
                         vegeta{\c{c}}{\~a}o",
               school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
                 year = "2000",
              address = "Sao Jose dos Campos",
                month = "2000-05-08",
             keywords = "vegeta{\c{c}}{\~a}o, Mato Grosso Do Sul (MS), espectr{\^o}metro 
                         imageador aerotransportado do vis{\'{\i}}vel e infravermelho, 
                         AVIRIS, {\'{\i}}ndice de vegeta{\c{c}}{\~a}o, analise 
                         espectral, contraste de imagens, infrared spectrometers, 
                         vegeetative index, spectrum analysis, image contrast.",
             abstract = "Investigou-se a influ{\^e}ncia do posicionamento e da largura das 
                         bandas do Vermelho (V - 600 a 700 nm) e do Infravermelho 
                         pr{\'o}ximo (IVP - 750 a 1100 nm), assim como a influencia da 
                         atmosfera, na determina{\c{c}}{\~a}o dos {\'{\i}}ndices de 
                         vegeta{\c{c}}{\~a}o (IVs) NDVI, SAVI e ARVI, obtidos a partir de 
                         dados hiperespectrais do sensor Airborne Visible/InfraRed Imaging 
                         Spectrometer (AVIRIS). Com base no uso do c{\'o}digo MODTRAN, os 
                         dados originais de radi{\^a}ncia aparente foram convertidos para 
                         imagens {{"}reflect{\^a}ncia} de superf{\'{\i}}cie\{"} e 
                         {{"}reflect{\^a}ncia} aparente\{"}. Os IVs foram obtidos para 
                         esses dois conjuntos de dados, considerando-se tr{\^e}s espectros 
                         de referencia representativos dos alvos Vegeta{\c{c}}{\~a}o 
                         Verde (VV), Vegeta{\c{c}}{\~a}o N{\~a}o-fotossinteticamente 
                         ativa (VNFA) e Solo (S), selecionados a partir da 
                         aplica{\c{c}}{\~a}o sequencial das t{\'e}cnicas Minimum Noise 
                         Fraction (MNF)e Pixel Purity Index (PPI). O objetivo foi 
                         identificar para cada IV o par de bandas V-IVP que possibilitasse 
                         o melhor contraste entre a VV e os demais componentes da cena. 
                         Foram utilizadas larguras de bandas estreitas (10 nm)e de bandas 
                         amplas simuladas (20 e 50 nm para o V, e 100 e 200 nm para o IVP). 
                         Tamb{\'e}m foram simuladas as bandas dos sensores ASTER/Terra, 
                         AVHRR/NOAA 14, MODIS/Terra, MOMS-2, MSS/Landsat 5, HRVIR/Spot 4 e 
                         TM/Landsat 5. Finalmente, correla{\c{c}}{\~o}es entre os IVs e 
                         os valores da fra{\c{c}}{\~a}o VV, extra{\'{\i}}dos de um 
                         modelo linear de mistura espectral, em fun{\c{c}}{\~a}o do 
                         posicionamento de bandas, foram obtidas e comparadas com 
                         resultados de um experimento radioel{\'e}trico de campo feito por 
                         Gleriani (1994). Os resultados mostram que a largura das bandas 
                         n{\~a}o influencia o contraste espectral entre os membros de 
                         referencia, quando s{\~a}o utilizados dados de reflect{\^a}ncia 
                         de superf{\'{\i}}cie, desde que n{\~a}o seja invadida a 
                         regi{\~a}o de borda vermelha (de 690 a 750 nm). Ao 
                         contr{\'a}rio, mudan{\c{c}}as no posicionamento das bandas, 
                         tanto do V quanto do IVP, provocam varia{\c{c}}{\~o}es no 
                         contraste entre os alvos. O contraste e otimizado com o 
                         posicionamento da banda do V em 677 nm, tanto para dados de 
                         reflect{\^a}ncia de superf{\'{\i}}cie quanto aparente, e para 
                         todos os {\'{\i}}ndices avaliados. Para a regi{\~a}o do IVP, o 
                         NDVI apresenta otimiza{\c{c}}{\~a}o do contraste com o centro da 
                         banda pr{\'o}ximo a 769 nm, mas o ARVI e o SAVI mostram 
                         varia{\c{c}}{\~o}es no posicionamento das bandas de melhor 
                         contraste, dependendo dos valores utilizados para as suas 
                         constantes y e L. A influ{\^e}ncia da atmosfera causa 
                         diminui{\c{c}}{\~a}o no contraste da cena, principalmente quando 
                         a largura da banda abrange bandas de absor{\c{c}}{\~a}o 
                         atmosf{\'e}rica no IVP. Os sensores MOMS, TM, ASTER, MODIS, HRVIR 
                         produzem melhor contraste do que o AVHRR e o MSS. Entretanto, o 
                         MODIS sofre menor influencia atmosf{\'e}rica, por possuir a banda 
                         do IVP mais estreita e melhor posicionada que os demais. Embora os 
                         resultados da correla{\c{c}}{\~a}o entre os {\'{\i}}ndices e a 
                         fra{\c{c}}{\~a}o-VV n{\~a}o sejam inteiramente concordantes com 
                         os do experimento radiom{\'e}trico de campo, eles revelam a 
                         tend{\^e}ncia do par de bandas V-IVP de melhor contraste 
                         espectral produzir tamb{\'e}m melhores estimativas dos 
                         par{\^a}metros da vegetacao. ABSTRACT: The influences of the 
                         positioning and width of red (R = 600-700 nm) and nearinfrared 
                         (NIR = 750-1100 nm) bands on the determination of AVIRIS-derived 
                         vegetation indices (Vls = NDVI, SAVI, and ARVI), as well as the 
                         atmospheric effects, were investigated. By using the MODTRAN code, 
                         apparent radiance AVIRIS data were converted into surface and 
                         apparent reflectance images, respectively. The Vls were calculated 
                         from both dataset, land for three end member spectra 
                         representative of green vegetation (GV), nonphotosyntetic 
                         vegetation (NPV) and soil (S). These endmember spectra were 
                         selected from the image itself through the sequential use of the 
                         minimum noise fraction (MNF) and pixel purity index (PPI) 
                         techniques. The objective was to identify the best RNIR band pair 
                         to optimize the contrast between GV and the other scene 
                         components. Narrow (10 nm) and broad bands (R with widths of 20 
                         and 50 {nm;} and NIR of 100 and 200 nm) were simulated in the 
                         present investigation as well as spectral intervals equivalent to 
                         some broadband sensors (ASTER/Terra, AVHRR/NOAA 14, MODIS/Terra, 
                         MOMS-2, MSS/Landsat 5, HRVIR/Spot 4 and TM/Landsat 5). Finally, 
                         correlations between Vls and fraction values of VV, derived from a 
                         linear spectral unmixing model, were calculated and compared with 
                         results produced with the use of a field spectroradiometric 
                         dataset collected by Gleriani (1994). The results of the use of 
                         surface reflectance data show that bandwidth is not an important 
                         factor to affect spectral contrast if the R and NIR bands are not 
                         inserted into the red edge domain (690-750 nm). On the other hand, 
                         shifts in R-NIR band placement produce significant changes in 
                         contrast. The best results were obtained with the R band 
                         positioned around the chlorophyll absorption interval (677 nm), 
                         and the NIR band placed at shorter wavelengths (769 nm). However, 
                         the results can vary in the NIR interval for ARVI and SAVI 
                         depending on the y and L parameters used in their calculation. 
                         Atmospheric effects reduce the spectral contrast of the scene, 
                         especially when the NIR bands are placed at atmospheric absorption 
                         band intervals. From the broadband sensors investigated, MOMS, TM, 
                         ASTER, MODIS, HRVIR produce better spectral contrasts between GV 
                         and other scene components than AVHRR and MSS. However, if one 
                         considers both, atmospheric effects and contrast, MODIS presents 
                         the best results because of the positioning of its narrow NIR band 
                         in a very convenient atmospheric window. Although there is no 
                         complete agreement between the correlations of the Vls with the GV 
                         fraction values (unmixing model), and with the canopy parameters 
                         (Gleriani's experiment), the results indicate that the best R-NIR 
                         band pair to enhance contrast tends to be also the best estimator 
                         of the vegetation characteristics.",
            committee = "Galv{\~a}o, L{\^e}nio Soares (orientador/presidente) and Novo, 
                         Evlyn M{\'a}rcia Le{\~a}o de Moraes and Ponzoni, Fl{\'a}vio 
                         Jorge and Epiphanio, Jos{\'e} Carlos Neves and Meneses, Paulo 
                         Roberto",
           copyholder = "SID/SCD",
         englishtitle = "Influence of remote sensor band positioning and bandwidth and of 
                         atmospheric efects on the vegetation index determinations",
                label = "8840",
             language = "pt",
                pages = "179",
                  ibi = "6qtX3pFwXQZ4PKzA/o5LHo",
                  url = "http://urlib.net/ibi/6qtX3pFwXQZ4PKzA/o5LHo",
           targetfile = "publicacao.pdf",
        urlaccessdate = "2024, Dec. 11"
}


Close