Doppler radar target recognition by neural networks

Erdoğan, Alpay
In this thesis, automatic recognition of Doppler radar targets is studied. The Doppler signal characteristics enabling target classification are discovered. The main intention is to identify the factors affecting target characteristics such as target speed, target aspect angle, target range and radar system effects. Proper preprocessing and normalization techniques are explored and applied to obtain invariant features against those factors. Extracted features are analyzed for class separability through clustering and classification techniques. The resulting feature vectors are classified by using basic classification methods and neural networks. For all stages of processing, from preprocessing to classification, the main idea is to iiiobserve the trade-offs in features and techniques with respect to classification rate, computing performance and realizability in real systems. For this study, the radar is a ground surveillance Doppler radar and the targets are moving man, helicopter, tank, car, bus, and truck.
Citation Formats
A. Erdoğan, “Doppler radar target recognition by neural networks,” Ph.D. - Doctoral Program, Middle East Technical University, 2002.