Measure of Realness for Target Image Scoring (MORTIS)

One obstacle to the increased use of smart weapons and unmanned air vehicles is a lack of appropriate data to train current or contemplated onboard automatic target recognition (ATR) systems. Generating synthetic data is an appealingly simple solution, but this requires a simulator that can generate data (image/non-image) of sufficient fidelity to ensure correct ATR training. Unfortunately, synthetic and real data can vary widely, and there is no data validation mechanism that determines when synthetic data is sufficiently “real” to serve as training data for an ATR that will classify real data. Measure of Realness for Target Image Scoring, or MORTIS, is an automated data validation system that identifies data that can be used for ATR training by scoring data “realness”. MORTIS uses a predictive approach to data classification, using many features and stepwise discriminant analysis. MORTIS attempts to learn to distinguish between real and synthetic data during a training phase and subsequently predicts if new data are real. When MORTIS identifies real data correctly, the synthetic data looks different than the real data, but when MORTIS fails to distinguish between the two types of data, the synthetic data is sufficiently real for ATR training.