Abstract-In order to make an accurate estimation of how much
rice has been planted, it is essential to calculate the area of rice
fields as precise as possible. Therefore, to delineate the edge of
rice fields and remove the area occupied by the widened ditch
roads around the rice fields caused by corner reflector effect in
SAR images, have become a critical topic in rice monitoring and
yield estimation using SAR images. To detect the linear features
of rice fields for rice monitoring, this paper introduces an edge
detection and extraction scheme in Radarsat SAR images by
applying a two-dimensional (2-D) filter in eight directions. To
reduce the isolated speckle noise on SAR images, an σ-filter is
performed firstly, and then the two-dimensional filter in eight
directions can be selected and applied to detect edges. Finally, a
nonlinear operator is used to threshold its magnitude and
produces the edge maps of Radarsat SAR images. The proposed
scheme achieves positive results of extracting linear features,
reducing the speckle noise effectively, and meeting the need of
practical radar remote sensing application in rice monitoring
and yield estimation.