Deinterleaving of radar signals with connected component labeling based clustering

This thesis aims to deinterleave pulsed and continuous wave signals efficiently and accurately with the proposed innovative clustering for the signal deinterleaving process, in the radar warning receiver. Military airborne platforms have radar warning receiver architectures to detect threat radar signals. It is essential to classify and determine the capabilities of threat radar sources for military aircraft during combat. In this scope, an innovative method is proposed that includes a computer vision-based connected component labeling method for efficient clustering of pulse and Frequency Modulated Continuous Wave (FMCW) radar signals. During clustering, it is represented solution with the signal separation inputs such as the time of arrival, the angle of arrival, the radio frequency, and the pulse width parameters. After clustering, inside each cluster, the radio frequency, bandwidth, sweep time, and chirp direction parameters are estimated for FMCW signals while Pulse Repetition Interval (PRI) information is extracted for pulse radar signals. The deinterleaving tests have been comparatively realized with simulations by using various radar signal sets. According to simulation test results, our proposed clustering method provides faster and higher accuracy performance when compared to the hierarchical clustering method. In addition, fixed and staggered PRI information is extracted with high performance, while FMCW signal parameters are estimated with a low error rate.