Vertex-oriented Metaheuristic for Radar Lobe Detection in Electronic Warfare Context

Deinterleaving is the task in electronic warfare aiming to sort pulses emitted by multiple radars. Most deinterleaving methods search temporal patterns or cluster pulses according to their primary parameters, yet rarely exploit pulse amplitude. In this article, we propose a lobe detection method for noisy observed scenes. It is useful for deinterleaving because it provides information about a subset of pulses, i.e. pulses from the same lobe have likely the same emitter. Our method first estimates a lobe vertex, then uses simulated annealing optimization to find the best parabola fitting the lobe. It is evaluated on some real observed scenes, and results show it successfully performs lobe detection even in case of a high outlier rate.