Desenvolvimento metodológico para a integração de dados de sensoriamento remoto multi resoluções para a representação da cobertura da terra em campos contínuos de vegetação e classificação por árvores de decisão / Methodology development for the multi resolution remote sensing data integration to the representation of earth covering in vegetation continuous filds and classification for decision tree

AUTOR(ES)
DATA DE PUBLICAÇÃO

2005

RESUMO

Now, countless efforts to quantify the monitoring of the changes of the covering of the earth are had based on analyses of data of the satellite Landsat, consuming for this, an expressive number of hours in laboratories for the processing of these data. Besides, estimates of different analyses (data of smaller resolution - AVHRR) have been generating conflictual results, caused, above all, for the differences of definitions and methodologies applied by the countless researchers in this area. This way, this work had as general objective to develop a methodology for the dynamics monitoring of the vegetable covering in Amazon, where a larger emphasis will be given in the integration of different sensor data and space resolutions (terra/MODIS, Landsat/TM and ETM+ and CBERS 2/CCD). So much, was used for the systems development, initially as base, the technique denominated Vegetation Continuous Fields (VCF). In that technique, the forest is described as percentage of arboreal covering, where the applied steps are obtained through an automated procedure to generate a regression with data of obtained trainings ( sensor TM, ETM+ and IKONOS), coincident with the data of wide space resolution of MODIS. As study area, it was chosen an area of Mato Grosso State , for presenting great representativity in relation to the Amazonian Brazilian in terms of earth covering. The methodological process consisted of establishing an exploratory analysis of the technique VCF and the development of the monitoring system, using the program ENVI and your corresponding language IDL. The elaboration of the methodological process happened, jointly, with the application of several tests, being used, for so much, different training data (obtained through different applied processes to the scenes ETM+, TM and CBERS-CCD), so much separately, as together. With this, the viability was verified (acuracity, larger automation and smaller time of processing) of the created model. This stage was complemented with a comparison of the best generated product (MODIS 500m) for the State of Mato Grosso, with the one of the PRODES Project, for the year of 2002. As result was obtained a system with capacity of obtaining data (maps), so much of 500m, as 250m of MODIS, in a period that can vary from one month to one year. The model presented satisfactory results when just using a training data, considering that the same has high representativity of your classes in relation to the MODIS data. Through your preliminary results and your comparison with the data obtained with the PRODES Project (year of 2002), the viability of the presented methodology was verified, making possible, still, a considerable earnings in the automation of the methodological process and, above all in the time of processing (approximately two hours) of your data. It is waited with this methodology that the monitoring of the dynamics of the vegetable covering aids, especially, the SIVAM Project, and also, complement projects already developed for the Brazilian Amazonian, as it is the case of PRODES.

ASSUNTO(S)

modis (radiometria) regression tree remote sensing sensoriamento remoto classificação satélites campos contínuos de vegetação remote sensors classifications modis (radiometry) sensores remotos satellites região amazônica amazon region Árvore de regressão vegetation continuous field

Documentos Relacionados