A MEASURE OF THE INFORMATION GAIN ATTRIBUTABLE TO CUEING

The high false alarm rate of automatic target recognition systems used in conjunction with imaging systems has precluded their use as completely autonomous devices for the targeting and launching of weapons in military systems and has prompted their use primarily as an aid to final target recognition by a trained observer. Automatic target recognition systems used in this manner are termed cuers and are assumed by many to improve the probability of detection and recognition of a target by an observer searching an image. The efficacy of these cuers and methods for evaluating their performance are under investigation since no quantifiable measure exists. The paper addresses the automated part of the problem and presents a proposed objective measure of cuer effectiveness based on an information measure. The usefulness of this measure is in discriminating among competing cuer designs. Cuers are classified according to the type of information that they extract from the image and a measure of information for each type of cuer is proposed.