Radar Echoes Classification Based on Local Descriptor

The use of radar in meteorology is in a permanent development, where the data collected are subject to many undesirable echoes. Nonetheless, we propose to process these lasts using Local Binary Pattern (LBP) descriptors and others variants that we combined to Support Vector Machine (SVM) classifier. The approaches were validated on images recorded in Bordeaux (France) by incoherent radar. Different neighborhood's sizes are explored, the results reveal that the variance (VAR) gives the best preservation of precipitations using radius 1 where 82.99% of ground echoes are suppressed when LBP
^riu2
are combined to VAR. The best identification is then obtained with simply the VAR.