Sensitivity of C-Vv and C-Vh Polarizations for Edge Extraction by Mathematical Morphology in a Deforestation

Tropical forests, like the Amazon, represent important reserves of biodiversity. However, processes such as deforestation have led to a reduction in vegetation cover. The objective of this study was to perform the extraction of edges, in a polygon of deforestation contained in a scene of the Sentinel Satellite 1, using the technique of Mathematical Morphology, in the Amazon rainforest, in the state of Mato Grosso, Brazil. The methodology was developed to obtain the external, internal and gradient edges of the shallow cut feature, by implementing arithmetic operations of dilation and erosion in the radar image in the VH and VV polarizations. Images from the Sentinel 2 satellite were used on the same acquisition date as the SAR images, spatially comparing the edges obtained by the VH and VV polarizations. As a result, the gradient border presented an outline with a smaller noise and spacing in relation to the external and internal borders. Areas where there were abrupt textural changes presented well-defined edges and without the presence of noise for the polarizations. On the other hand, in locations with greater signal variability, such as agricultural regions, edges characterized by dispersed segments and high noise distribution were observed. Even so, the edges generated by the polarizations did not present significant differences between them. However, the edges obtained from the VH polarization stood out for presenting a smaller difference and higher (R²) in relation to the reference product. This suggests that edge extraction based on cross-polarizations can be an accurate approach for analyzing disturbances in tropical forests, such as in deforestation monitoring.