The USC Andrew and Erna Viterbi School of Engineering USC Signal and Image Processing Institute USC Ming Hsieh Department of Electrical and Computer Engineering University of Southern California

Technical Report USC-IPI-970

“High Resolution Digital Radar Imaging of Rotating Objects”

by Yeh-Hua Peter Chuan

June 1980

This dissertation is devoted to the imaging aspect of the problem of obtaining high resolution images of practical radar targets with digital processing techniques. The motion compensation aspect of the problem is also briefly described. A multifrequency stepped (MFS) radar is assumed and the Fourier transform relationship between the data (dimensioned in aspect angle and signal frequency) and the target reflectivity function is derived in both 2-D and 3-D forms. Assuming that the data is available for 360 degree aspect angle and using wideband radar, a coherent digital processing method is developed which will give the best possible resolution. Such a situation occurs when the target makes a complete turn. It is found that for such an imaging system the resolution is inversely proportional to the mean carrier frequency if such frequency is large compared to the signal bandwidth. In the case when the data is undersampled in range or aspect angle or both, a modified coherent digital signal processing technique is described that will get around such difficulty. It is found that the modified processing method gives poorer resolution but is better than either the mixed processing method or the incoherent processing method. That latter two processing techniques are also described in this dissertation. Experimental results are also presented and problems with real targets such as shadowing, glint and scintillation are discussed.

In the 2-D case, the radar data are sampled in polar coordinate format. The sampling requirements in this sampling scheme are discussed in great detail. Results from Doppler processing and Degrees of Freedom concepts both show that polar coordinate sampling in the Fourier domain is adequate if the inverse of the greatest sampling interval (in either radial or cross-radial dimension) in the Fourier domain covers the entire linear extent of interest in the real domain. Analytical methods using Poisson's summation formula show the same results in more detail especially in predicting undersampling effects. The results on polar coordinate sampling can be applied to other systems in which polar format sampling is a natural setting.

To download the report in PDF format click here: USC-IPI-970.pdf (11.0Mb)