Identification of Moving Ground Targets by a Low Cost, Portable Radar System

Abstract

Identification of targets with measurements taken from low-cost, portable, ultra-wideband radar system was carried out using 2 different methods based on range profiles or timefrequency representations. In order to eliminate the clutter in the data set, the component reflected from the target in the measurement was considered sparse and all the remaining components were considered low rank, and the target and clutter components were separated. Clutter-free range profiles or time-frequency representations generated from the corresponding frequency domain data are classified with K-nearest neighbor (KNN) and the targets are identified. It has been shown that the identification based on time-frequency representations is less affected by clutter than the method based on range profiles, and its performance is higher because it includes phase information.