Combination of Support Vector Machine and H-Alpha Decomposition for Subsurface Target Classification of GPR

Abstract

Subsurface target classification of GPR is a hot topic of geophysical field which is aimed to classify different kinds of targets based on their attributes, such as polarimetric attributes and geometrical features, although the existing methods can classify different targets, but they are not efficient and intelligent enough, especially in dealing with data of large amounts. Support Vector Machine is a method of Machine Learning which is used to classify different kinds of samples based on their attributes. We combine Support Vector Machine(SVM) with H-Alpha Decomposition for subsurface target classification of GPR. We use H and $\alpha$ as parameters of SVM for target classification. To test the effect of the combination of these two methods, we process the polarimetric data of three different kinds of targets measured in laboratory and obtain the data of H and $\alpha$, then we use the data of H and $\alpha$ to test the support vector machine and it turned out to be effective and feasible, and the accuracy is relatively high.