Internet of target perception event driven based on machine vision using a fast extraction algorithm

With the complexity and diversity in real-time recognition and tracking of target in the complicated electromagnetic environment in mind, the paper proposes a target recognition and tracking method based on scale invariant feature by dint of the technology of image registration to effectively reflect the features distribution of target image, thereby improving the reliability of tracking and recognition system. The experiment shows that this method can effectively reduce errors of tracking, precisely recognize target position and improve target tracking precision when the target changes with large angle and the background transforms violently.