Example-based object detection in images by components

In this paper, we present a general example-based framework for detecting o bjects in static images by components. The technique is demonstrated by dev eloping a system that locates people in cluttered scenes. The system is str uctured with four distinct example-based detectors that are trained to sepa rately find the four components of the human body: the head. legs, left arm , and right arm. After ensuring that these components are present in the pr oper geometric configuration, a second example-based classifier combines th e results of the component detectors to classify a pattern as either a "per son" or a "nonperson." We call this type of hierarchical architecture, in w hich learning occurs at multiple stages, an Adaptive Combination of Classif iers (ACC). We present results that show that this system performs signific antly better than a similar full-body person detector. This suggests that t he improvement in performance is due to the component-based approach and th e ACC data classification architecture. The algorithm is also more robust t han the full-body person detection method in that it is capable of locating partially occluded Views of people and people whose body parts have little contrast with the background.