Border surveillance is one of the top security priorities for countries. It has been proven that the classical method does not provide complete security. An effective solution is the use of wireless sensor networks (WSN) to improve border surveillance system. In surveillance application scenario based on WSN, the detected target may be an unarmed person, a soldier carrying ferrous weapon or a vehicle. So that, after detection, the system must identify target type, in order to be tracked at later if required. This work focuses on classification method where we introduce our classification approach not too resource-intensive, easy to implement and suitable for the large-scale environment, using probabilistic techniques and supervised learning methods to address multi-sensing data correlation