SBIR-STTR Award

Although the affects of sensor parameters on human recognition performance have been studied extensively, work done to relate sensor parameters to automatic target recognition (ATR) system performance is still in its infancy. This proposal outlines a program with the following objectives: (I) develop a computational method for modeling atr system performance which enjoys the same characteristics the mrt measure does for man in the loop performance modeling. This method will be based on measuring how each component of an image processing system distorts the shape of a target or clutter object. (II) find the functional relationship between various candidate measures of shape distortion and the probability of recognizing a target. (III) find a parametric set of test patterns which is homomorphic to real targets and clutter with respect to the chosen shape distortion measure so that target and clutter independent performance results may be achieved by analyzing only a small set of data.

Keywords:
image processin target recognit performance(mec computer archit algorithms pattern recogni