Automated image-based range performance evaluation using TOD over natural backgrounds

Abstract

The use of a synthetic observer model has shown promise for range performance analysis of novel imaging systems. This has many advantages over traditional analytical range models, chiefly stemming from the fact that it determines performance from (real or simulated) imagery directly, rather than from a pre-specified list of parameters. Our synthetic observer approach operates over a Triangle Orientation Discrimination (TOD) target and observer task, using a template correlator for target identification. The synthetic observer performance is taken as a proxy for human target identification performance, enabling expedient evaluation of image processing pipelines, sensor configurations, environmental conditions, etc. In prior work we have explored how the template-correlator-based synthetic observer performs on flat background, flat target imagery. In this work, we apply the same synthetic observer design to natural backgrounds. Performance is compared to that of human observers on the same perception task.