Aimed at the problem of the fast clustering sorting of radar signals under complex conditions of electromagnetic environment, a fast clustering sorting method of radar signals is proposed by adopting a variable precision rough set model based on the similarity of border pulse. In view of some reasons that the similarity of the weighting matrix computed by the original method is determined by experts or experience on no consideration of radar radiated properties, this leads to a low clustering sorting accuracy rate. The improved method is used to obtain the weights of feature parameters in radar emitter by utilizing the variable precision rough set model, further compute similarity by the weighting matrix composed of optimum feature weights, and then perform fast clustering sorting of radar signals by using similarity of border pulse. The simulation results show that the improved method increases the clustering sorting accuracy rate by 5.93% under the same fast clustering sorting capability compared with the original method.