The detection of buried landmines or unexploded ordnances is a problem of great interest in both civilian and military applications. In such a framework, this paper presents an approach for real-time localization of unknown subsurface objects. Such a technique is aimed at defining high resolution risk maps by means of a classification strategy based on a support vector machine and a multi-resolution approach. In order to assess its effectiveness, a selected set of numerical results regarding multiple target scenarios are presented.