Object/Target Discrimination, Recognition, and Identification

Automatic target recognition independent of range and orientation is achieved by using a very different type and much reduced quantity (fewer pixels) of visual information, easily derived from the output of any standard image sensor by variance sub-sampling. The very low pixel density of the resulting variance image directly provides (1) high pose-immunity in any one target-reference match, and (2) enables multiple (variance) references to be stored of each target type to provide coverage over the full set of required poses. In spite of its low pixel density, variance sampling provides high resolution information essential for recognition. Conventional Nyquist sampling, on the other hand, requires high pixel density to provide the essential high resolution information and is thereby highly pose-sensitive and also highly processing and memory intensive. This very different (variance) and much reduced quantity of visual information is used almost exclusively by human vision to perform recognition. Its transition to ATR systems, that seek to emulate human vision, will result in a major advance in ATR capability.