Passive sensors or radars in a jammed environment can only measure the targets' bearing or direction information and their location then can be calculated by using well known triangulation methods. However triangulation produces ghosts which act and operate like real targets. Although there are methods in the literature for ghost elimination the procedures proposed are usually very complex especially if the measurements are noisy. Moreover, the resulting false alarm rate might be unacceptable. As a way of addressing this problem, in this paper, a new efficient ghost elimination algorithm is provided. The proposed approach uses the fact that a ghost trajectory is a function of the trajectories of two targets and, hence, its complexity is higher. The algorithm estimates the complexity of the observed trajectories (order of systems that approximately generate them) by using fast algorithms based on the concept of atomic norm and classifies observed objects as targets/ghosts if their trajectory complexity is low/high. Finally the proposed approach is illustrated by using an academic example.