Say I would like to identify and bounding-box-track a person within a given video. But if there is the advantage that I have prior knowledge of what the person looks like -- e.g., their body features, their clothes, etc. (i.e., extensive offline training possible), then it seems like I should be able to exploit that to make the tracking more accurate and drift-free.
In the literature, is there a name for the class of CV algorithms/research that is concerned with this slightly different problem (as opposed to standard tracking of a previously unknown target)?