Automatic Target Recognition

Existing automatic target recognition (ATR) systems require retraining whenever the operational environment changes or new classes of targets need to be added. Such offline retraining usually causes operational downtime. Sponsored by U.S. Navy, UC has developed an Incremental Knowledge Assimilation System (IKAS), a novel ATR software toolkit for detecting and classifying underwater mines from sonar images of the sea floor. IKAS uses the latest advances in deep learning. Capitalizing on its unique capability to adapt to varying operational environments, IKAS has the following benefits:

  • Eliminates operational downtime while training.
  • Improves efficiency and effectiveness of the mine hunting operational process.
  • Minimizes human intervention and associated cost while maintaining an equal or improved recognition performance.