Application of a Decomposed Support Vector Machine Algorithm in Pedestrian Detection from a Moving Vehicle

摘要

For a shape-based pedestrian detection system, the critical requirement for pedestrian detection from a moving vehicle is to both quickly and reliably determine if a moving figure is a pedestrian. This can be achieved by comparing the candidate pedestrian figure with the given pedestrian templates. However, due to the vast number of templates stored, it is difficult to make the matching process fast and reliable. Therefore many pedestrian detection systems re developed to help the matching process. In this paper, we apply a decomposed SVM algorithm in the matching process which can fulfill the recognition task efficiently.