Moving Target Simulator (MTSim) Framework

Airborne Moving Target Indication (AMTI) from proliferated Low Earth Orbit (LEO) requires rapid detection and identification of moving targets from a sequence of imagery which contains platform induced motion, natural scene motion and target motion. Successful AMTI in such a complicated and cluttered environment requires new and innovative image processing techniques. Machine Learning (ML) may provide creative solutions in this domain but requires a large training set of images, which represent the full range of possible targets, backgrounds, and viewing scenarios.  Obtaining such training sets is not possible from measurement alone because operationally important scenarios may be absent or limited. Radiometrically accurate synthetic data can expand available training datasets to improve algorithm performance by providing large quantities of automatically labeled training data for rare scene events and mission critical non-ideal conditions. To address the need for the additional training imagery, Spectral Sciences, Inc. (SSI) proposes to develop the Moving Target Simulator (MTSim) framework for the simulation of time sequence synthetic imagery with embedded moving targets. MTSim combines state of the art 3D radiative transfer scene simulation with a sensor response module to enable the Air Force to produce accurate synthetic moving target imagery as measured by any on-orbit system.