SBIR-STTR Award

In this project FASTMAN will determine the feasibilityof using fractal and wavelet signal preprocessing in conjunctionwith neural network pattern classifiers on high-range resolutionradar data for detecting cold stationary ground targets in aclutter-rich environment. Previous researchers have demonstratedthat measurement of the fractal dimension of high-rangeresolution (HRR) returns can be used to distinguish stationarytargets in clutter rich environments. Other groups of researchershave shown that the Wavelet Transform can be used to accuratelyand efficiently measure the fractal dimensions of signals byexamining the behavior of the transformed signals across scales. We propose to marry these two concepts into a signal preprocessorin which the Wavelet Transform will compute the local fractaldimensions of HRR returns, and those fractal dimensions will forma feature vector which will be used by neural network patternclassifiers to distinguish targets in the backscattered signal.We will also investigate the applicability of two other Waveletprocessing techniques for radar signal processing: the use of theAdaptive Wavelet Transform to extract key target features and theuse of doppler-tolerant wavelet radarpulses.

Detecting stationary targets in a ground clutter background is a difficult problem for airborne radar sensors. Traditionally, adaptive amplitude thresholding is employed (e.g. constant false alarm rate (CFAR) processing). However, when using radar amplitude processing techniques, the desired high probabilities of detection can only be achieved by setting thresholds very low, often resulting in an unacceptable number of false-alarm targets due to clutter. Increasing those thresholds to produce an acceptable number of false alarms results in an unacceptably low detection probability. Fractal geometry techniques, in which the numerical value of the fractal dimension of radar signatures are used to distinguish man-made "regular" objects from "natural" clutter, have recently been developed and have shown a great deal of promise. In Phase I, FASTMAN used the relationship between the Wavelet Transform and fractals to produce a robust, computationally-efficient method to measure (estimate) the fractal dimension of radar signatures (and the corresponding target surfaces), even in the presence of significant noise. In Phase II, we will enhance our algorithm, extend its underlying theory, and test it on Longbow MMW radar data.