Incremental appearance model based target tracking using feedback resampling strategy

Target tracking has been widely applied in the field of computer vision and extremely promising. It is also one of research hot difficulties. To its difficulties derived from information loss of three-dimensional world projecting on two-dimensional images, target appearance texture and the change of the geometry, the light changes, local or global screening etc. Incremental subspaces through online learning mechanism, at a fixed frequency update appearance model, can track the target in time series and capture target appearance changes in time and space. Although traditional IVT can well adapt to illumination changes and the appearance of the target, intense movement and occlusion occurs on the space fail to tracking. To this end, we put forward a kind of feedback learning algorithm based on incremental formula of space in the framework of particle filter, the introduction of true observation samples the verdict as feedback information, provide the basis of updating subspace and a particle filter resampling.