Target Identification Algorithm for the AN/AAS-44V Forward Looking Infrared (FLIR)

Accurate identification of unknown contacts is a crucial issue in military intelligence. In order for this task to be accomplished by human observers, each one must be specially trained and regularly re-qualify. Even with training, their decisions are subject to human error: bias, expectations, or even a lack of sleep may compromise their accuracy. If an automated system could quickly and accurately determine the identity of a contact, it would be a great benefit. This thesis explores some of the problems which must be addressed in producing such a system. We detail an approach to an algorithm which compares a picture of an unknown ship to an established database and determines its most likely classification. In particular, we use infrared images from FLIR video taken at sea, and obtain classification results for a small test set. We tested eighteen images with success rates varying between seventy-three and eighty-nine percent.