摘要：The concept of spectral mixture offers a wide range of applications in the Remote Sensing area. The application of this concept, however, requires the prior estimation of the component’s (endmembers) spectral response. This latter requirement can be achieved by different methods, as reported in the literature, such as techniques for the detection of pure pixels, use of spectral libraries, and field radiometric measurements. Among those, the most often used is the pure pixel approach. In this approach, the components’ spectral reflectances are estimated by means of pixels covered entirely by a single component. This approach offers the advantage of allowing the extraction of the required spectral reflectance directly from the image data. This approach, however, becomes increasingly unfeasible as the spatial resolution of the image data decreases, due to the larger ground area covered by a single pixel. In this study we propose a methodology to estimate the spectral reflectance for each component class in moderate spatial resolution image data, by applying the linear mixing model (MLME), and higher spatial resolution image data as auxiliary data. It is expected that this methodology will provide a more practical way to implement the spectral mixture approach to moderate resolution image data, allowing in this way the expansion of the information about the components’ proportions across larger areas, up-scaling information in regional and global studies. Experiments were carried out using CCD (20 m ground resolution) and IRMSS (80 m ground resolution) and WFI (260 m ground resolution) CBERS-2 image data, as medium and moderate spatial resolution data, respectively. The spectral reflectances for the components in the IRMSS and WFI CBERS-2 spectral bands are estimated by applying the proposed methodology. The reliability of the proposed methodology was assessed by both analyzing scatter plots for CBERS-2 data and by comparing the fraction images produced by image data sets of the sensors analyzed.
其他摘要：The concept of spectral mixture offers a wide range of applications in theRemote Sensing area. The application of this concept, however, requires theprior estimation of the component’s (endmembers) spectral response. Thislatter requirement can be achieved by different methods, as reported in theliterature, such as techniques for the detection of pure pixels, use of spectrallibraries, and field radiometric measurements. Among those, the most oftenused is the pure pixel approach. In this approach, the components’ spectralreflectances are estimated by means of pixels covered entirely by a singlecomponent. This approach offers the advantage of allowing the extraction ofthe required spectral reflectance directly from the image data. This approach,however, becomes increasingly unfeasible as the spatial resolution of theimage data decreases, due to the larger ground area covered by a single pixel.In this study we propose a methodology to estimate the spectral reflectance foreach component class in moderate spatial resolution image data, by applyingthe linear mixing model (MLME), and higher spatial resolution image data asauxiliary data. It is expected that this methodology will provide a morepractical way to implement the spectral mixture approach to moderateresolution image data, allowing in this way the expansion of the informationabout the components’ proportions across larger areas, up-scaling informationin regional and global studies. Experiments were carried out using CCD (20m ground resolution) and IRMSS (80 m ground resolution) and WFI (260 mground resolution) CBERS-2 image data, as medium and moderate spatialresolution data, respectively. The spectral reflectances for the components inthe IRMSS and WFI CBERS-2 spectral bands are estimated by applying theproposed methodology. The reliability of the proposed methodology wasassessed by both analyzing scatter plots for CBERS-2 data and by comparingthe fraction images produced by image data sets of the sensors analyzed.
关键词：Mistura espectral;resolução espacial moderada;dados CBERS;Spectral mixture;moderate spatial resolution;CBERS Data
