In this paper, a two-step methodology is developed to deinterlace RADAR signals. We mainly worked from a data simulator to obtain a large diversity of signals representing typical RADARs. First, a clustering algorithm is used to separate the pulses in the F-PW (Frequency-Pulse Width) plane, then a phase of cluster agglomeration is performed using the cluster output of the previous step as a features for an hierarchical agglomerative clustering combined with optimal transport distances. Results on labeled simulated data are given.