| Popis: |
A spectrum of knowledge‐based systems has been designed and applied to target acquisition, classification, and tracking. A survey of these leads to the conclusion that more profound techniques need to be developed and applied. Particularly necessary are techniques that match the performance of a highly trained sensor suite operator. Preferably, such a system would function autonomously. A knowledge‐based system (called “PING”) for autonomously classifying and tracking underwater targets in terms of their acoustic signals as well as their context was designed. PING is based upon earlier work in the foundation of analogical reasoning. The performance of PING was tested and compared with the performance of a population of highly trained sonar operators; the system was also compared with one employing solely the Demster‐Shafer evidential reasoning method. In both tests, PING exceeded the performance of the test controls. |
