Automated EO/IR Detection Techniques For Floating Objects

We propose to further develop the computer vision system created in Phase I of this project, and to adapt it to the area of underwater mine detection. Adaptation to the new environment is anticipated to be relatively straightforward, as our neural network approach has been shown to generalize well to very different circumstances after retraining. This approach allows for natural integration of multiple signals, in this case, optical and sonar. This system will be optimized for use in the Archerfish underwater drone, and the associated remote user interface. The availability of relative motion from underwater drones will allow us to leverage our work in integrating visual information over time. We will engage in an architecture process that emphasizes both modularity and openness, such that our final product will be straightforward to integrate with components of the airborne mine neutralization system (AMNS) in the helicopter, but also adaptable by others for new uses.