Ground echoes filtering using the completed local binary pattern and the support vector machine

Meteorological data are subject to many spurious echoes, mainly due to ground echoes and anomalous propagation. The removal of the undesirable echoes is obligatory since they cause several errors in the estimation of precipitations. The proposed method is based on the combination of the Completed Local Binary Pattern (CLBP) which is textural descriptor and the Support Vector Machine (SVM). Two incoherent radars site are considered Setif (Algeria) and Bordeaux (France) to develop this method. The results obtained show that more than 98,2% of ground echoes are removed for Bordeaux and 89.4% for Setif while 92,3% of precipitations are preserved in Bordeaux and 97.95% for the region of Setif. In addition the anaprops that exist only in Bordeaux are suppressed with 91.38%.