High performance automatic target recognition

Designing a vision system, which was motivated by that of the human eye, has been done since the introduction of digital computing devices. Its computational complexity hinders it from the required accuracy and flexibility achievable by these systems. A real-time Automatic Target Recognition system having the ability of detection, identification and tracking of pre-specified targets even with bad visual artifacts is developed. This work investigates an improved combination of detection and recognition algorithms to reach a better solution to the problem of target detection and recognition while aiming to fit the design of a minimal system. A speed up of 2.4 is achieved, by using a non-preeminent GPU, over brute force implementations.