CARBON BALANCE AND RADIATION BALANCE IN DIFFERENT SOY PLANTING SYSTEMS BY SPECTRUM-TIME VARIABLES.
Agricultural culture, multispectral imaging, orbital products, data mining.
The objective of this research was to identify areas cultivated with soybean (Glycine max L. Merr.) in the state of Mato Grosso (MT) through multispectral images of MODIS (Moderate-Resolution Imaging Spectroradiometer) sensor and to develop an applied methodology for Plantation detection. Conventional and Direct, based on GEOBIA (Geographic Object-Based Image Analysis) and MD (Data Mining) in the processing of spectral bands and vegetation indices in a Landsat satellite time series for crop areas. For the identification of the cultivated areas with soybean cultivation, the PCEI (Perpendicular Crop Enhancement) was used and in the distinction of the type of tillage two combined techniques were used: i) GEOBIA and ii) MD. The spectral bands and vegetation indices were properly treated and a multiresolution segmentation was applied to define the objects, with the predefined selection of the targets, and a decision tree (AD) was generated by the C4.5 / J4 algorithm. In this algorithm, 468 sample objects were used to obtain a matrix error analysis and Kappa statistics. Statistics indicated that the rating achieved an overall accuracy of 95.29% and a Kappa coefficient of 0.92. The results revealed that the combination of GEOBIA and MD techniques are efficient and promising for the soil treatment type classification process.