Improving Joint Tracking and Classification with the Transferable Belief Model and terrain information

This paper presents an extension to an existing Joint Tracking and Classification approach for Wireless Sensor Networks. Terrain information and how it restricts target movement is used, along with the estimated target speed, to classify a target using the Transferable Belief Model. When using terrain information, a method is used to take into account the uncertainty of the target position. In addition, a mechanism is added to the Transferable Belief Model to prevent the assignment of belief to outcomes that past behaviour have shown is not possible. The improvements are analysed with the use of three simulated scenarios - the results of which show consistently good classification performance. In each scenario, targets have non-constant velocity trajectories.