Abstract: A computer-implemented method for probabilistically classifying an occurrence of an event, a change in the state of a target, includes measuring feature data of the target simultaneously processing the measured feature data through first and second filters. The first filter is suited for a situation in which the target is in a first state and generates a first residual and a first residual covariance for the measured feature data. The second filter is suited for a situation in which the target is in a second state and generates a second residual and a second residual covariance for the feature data. By determining a probability of the occurrence of the event and the probability of the non-occurrence of the event and comparing the two probabilities with at least one threshold value the occurrence or non-occurrence of the event is determined.

Abstract: A hostile missile engager senses the missile and supplies kinematic data to an interceptor missile fire control processor, which predicts the target's future location with the aid of a powered/unpowered identifier. The identifier passes the kinematic data through filters having lags for powered and unpowered operation, to produce residuals and residual covariances. The probabilities of powered and unpowered motion are determined, and thresholded. If the probability of powered motion or unpowered motion exceeds its threshold, the motion is deemed known. If the target is deemed to be powered, a set of the three-dimensional kinematic features is applied to a nine-state Kalman filter to produce an optimal state. If the target missile is unpowered, a set of the three-dimensional kinematic features corresponding to the second lag is applied to a six-state Kalman filter to produce an optimal state for the unpowered motion. The optimal states control the interceptor toward the target.