A cascaded classifier for pedestrian detection

In a pedestrian detection system, the most critical requirement is to quickly and reliably determine whether a candidate region contains a pedestrian. It is essential to design an effective classifier for pedestrian detection. Until now, most of the existing pedestrian detection systems only adopt a single and non-cascaded classifier. However, since the scene is complex and the candidate regions are too many (in our experiments, there are more than 40,000 candidate regions); it is difficult to make the recognition both accurate and fast with such a non-cascaded classifier. In this paper, we present a cascaded classifier for pedestrian detection. The cascaded classifier combines a statistical learning classifier and a support vector machine classifier. The statistical learning classifier is used to select preliminary candidates, and then the Support vector machine classifier is applied to do a further acknowledgement. This kind of cascaded architecture can take both advantages of the two classifiers, so the detecting rate and detecting speed can be balanced. Experimental results illustrate that the cascaded classifier is effective for a real-time detection.