EXO-ATMOSPHERIC DISCRIMINATION OF THRUST TERMINATION DEBRIS AND MISSILE SEGMENTS

This article explores a time-delay neural network (TDNN) for exo-atmos pheric discrimination of a missile reentry vehicle (RV) from other mis sile parts and thrust termination debris. The TDNN is an enhanced vers ion of a back-propagation neural network that accounts for the feature s in the time domain by using the rate of change of the infrared signa ture over several seconds as a discriminant. We used simulated infrare d signatures to train and test the TDNN on 90 randomly selected scenar ios. Results showed that the TDNN could identify the RV in 97% of the cases, for a leakage rate of 3%; the false alarm rate (percentage of c ases for which a non-RV was identified as an RV) was 5%.