2013 | OriginalPaper | Buchkapitel
Real-Time Multi-pedestrian Tracking Based on Vision and Depth Information Fusion
verfasst von : Shan Gao, Zhenjun Han, Ce Li, Jianbin Jiao
Erschienen in: Advances in Multimedia Information Processing – PCM 2013
Verlag: Springer International Publishing
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Visual object tracking plays an essential role in vision based applications. Most of the previous Multi-pedestrian Tracking has limitations due to considering each pedestrian with the same motion and appearance model in a uniform observation space, leading to tracking failures in complex occlusions. To address this problem without losing real-time performance, we propose a graph based approach for multi-pedestrian tracking using fused vision and depth data in this paper, where one main contribution is devoted in terms of the consideration of pedestrians with different priori probability in distinguishing observation space divided based on vision and depth information. Then we formulate the tracking model using an Improved Bipartite Graph (IBG), which is then optimized with a heuristic algorithm. Experiments on three datasets of fused vision and depth data demonstrate robust tracking results of the proposed approach.