| 摘要 |
A target correlation matrix is generated for multiple two-class combinations of target types each having a target correlation and a synthetic aperture radar observation space. A target probability density of a target radar cross-section signature and a background probability density of a background radar cross-section signature are utilized. The observation space of each of the two-class combinations is partitioned into a target partition and at least one background partition in accordance with the target correlation. A conditional log likelihood is calculated using at least one random number for each of the partitions in accordance with the target probability density and the background probability density, and summed according to the two-class combinations. A maximum log likelihood is calculated from the summed conditional log likelihoods given that one target type of the multiple two-class combinations is assumed to be true. An automatic target recognition performance prediction based on the maximum log likelihood is generated.
|
