In general, a shape-based[1] pedestrian detection system includes the following two steps:
(a) finding out and tracking a possible pedestrian figure, and
(b) determining if the candidate pedestrian figure is really a pedestrian figure by checking if it matches with any of the
pedestrian templates.
Since there are a large number of templates, it is necessary to build up a search tree for the Match process [2,3]. Each node
in the tree is one feature of the corresponding templates that can be used for classification and where each branch is one
pedestrian template. Usually, the search tree is not adjustable during the matching process.