SVM Classifier for Meteorological Radar Images

This paper proposed an efficient approach, for weather Radar image classification and filtering, which is based on the combination of the textural analysis and support vector machine (SVM). Since the images are formed by two kinds of echoes, precipitations (rainfalls) and ground echoes (clutter), caused by earth's surface, thus, our purpose is to preserve precipitations and eliminate ground echoes from radar images recorded by weather radar of Setif. The filtering method consists of extracting features using gray level co-occurrence matrix (GLCM) for each type of echo. In order to identify data formed by the elements of the useful textural parameters, a Support Vector Machine classifiers were used with different forms of kernel function. The achieved result for the mean rate of correct recognition of echoes is about 94.77%. Keywords— SVM, Image, Radar, Precipitations, Ground echoes.