Abstract:

Radar signals are preselected by the clustering of incident pulses using their DOA (direction of arrival) parameters. In order to improve the clustering reliability, a novel method of calculating the similarity coefficients of DOAs is proposed, then the clustering number of the incident pulses is judged by the method of decomposing the eigenvalue of the similarity matrix which is built with the similarity coefficients, it is proved to be reliable that when the difference of scattering degrees is larger than 0.34, the mission success rate of the judged clustering numbers approaches 80% or larger. Finally, a hierarchical clustering method based on the clustering numbers judged above is analyzed. After being tested with multiple data, this method is verified to be simple and successful.