Pedestrian detection for intelligent vehicle based on bilayer difference features algorithm

In this paper, a new powerful pedestrian feature extraction algorithm named bilayer difference feature is proposed. In this new feature, two different features are used for detecting pedestrians and eliminating false alarms respectively. The upper layer of bilayer difference feature is constituted by improved accumulate binary Haar (ABH) feature. ABH feature could enhance the ability of illumination invariance and keep high real-time performance, which is used for fast pedestrian detection and location; while, Edgelet feature could be used for the lower layer of the bilayer difference feature and check the validity of pedestrian detected by ABH feature, reduce the influence of false alarms and occlusion to detection. The experiment carried out on videos from three famous datasets illustrates that the proposed method is real-time and feasible enough for pedestrian detection in intelligent vehicle environment.