Evaluation of a Laboratory Automatic Target Recognition system: Air-to-Sea ImageSequence

摘要

The operation and evaluation of a laboratory automatic target recognition (ATR)system is described. The ATR system was evaluated against its ability to acquire and track a specific target in the presence of false targets and background clutter. A ten second air-to-sea image sequence that contained the infrared signatures of three ship targets and sea clutter was used as the test sequence. The evaluation involved assessing the system performance with default values for various system parameters. These were then varied and the effect on system performance measured. The ability to acquire and track the true target over a limited range of target orientations and in the presence of low clutter and false targets was demonstrated. Furthermore, it was found that acquisition was easier if the target was somewhat broadside rather than end-on to the sensor. A shortcoming of the system was that it was not robust to minimal target obscuration. In addition, its ability to discriminate between the true and false targets was somewhat limited, and resulted in the acquisition and tracking of false targets. The evaluation showed that the system discrimination was improved by optimization of a number of parameters, however, this also resulted in a reduction of its ability to acquire and track the true target. In addition, those parameters that were most important in terms of their effect on the system performance.