Multi-Pedestrian detection and tracking using unified multi-channel features
Young Chul Lim, Min-Sung Kang · 2017
This paper presents an online multiple pedestrian detection and tracking method using unified multi-channel features. The proposed method efficiently utilizes the multi-channel features by sharing them in each module: pedestrian detection, visual tracking, and data association. The multi-channel features are originally generated from the pedestrian detection module, and they represent sufficiently rich feature information. In the pedestrian detector, feature vectors for the pedestrian classifier are constructed from the unified multi-channel features, and the visual tracker localizes the target pedestrian on the multi-channel feature maps. An appearance model for data association is also established from the unified multi-channel features. Experimental results show that our method outperforms the state-of-the-art method in both accuracy and speed.