Wavelet-based compression of HRR radar data for moving targets

Automatic recognition of moving targets has been a topic of much recent interest. In particular researchers have been considering the exploitation of Doppler-related features resulting from moving tracks, wheels, etc., to improve performance. In developing and testing such algorithms, data with three or more dimensions are often used resulting in very large data sets. We describe a system for compressing radar data sets with the dimensions of range, Doppler, and azimuth. Our approach employs smooth localized exponentials to model the scattering returns. A best-basis approach is employed to find an optimal wavelet transformation, and the resulting coefficients are entropy coded. We present results showing the effectiveness of our approach and compare to a more conventional Fourier-based technique.