Using Position Uncertainty in Recursive Automatic Target Classification of Radar Tracks

In this paper, we show how radar plot uncertainty can be included leading to a more robust classification of targets observed by a rotating 2D radar. Targets far from the radar will have a greater uncertainty in the position and therefore the estimated speed of the targets will be more uncertain. The uncertainty is sensor dependent and will therefore need to be taken into account when classifying based on training data from multiple different sources. Including the uncertainty in the radar plot positions, leads to an improved estimate of the probability of a target belonging to any given class in a list of possible classes. We show results for two synthetically generated cases, where we include the uncertainty and from a real world radar scenario.