Autonomous online learning method based on random fern classifier

The present invention discloses a fully autonomous online learning method based Random Ferns classifier, which can only be a goal box selected for the target class online learning classifier in the video frame. Step: first target marquee obtained using the affine transformation of the initial positive sample set, extracting a small negative initial training sample set is classified Random Ferns in non-target area of ​​the video; second, using the classifiers in the video frame Target Detection. The detection process, the nearest neighbor classifier learning to collect new samples online, and automatically determine the sample type; Finally, the new online training for random sample fern classifier update posterior probability random fern, a gradual increase in target detection classifier precision target detection system to achieve full self-learning.