作者: Monique Thonnat , François Bremond , Bernard Boulay
DOI:
关键词:
摘要: This paper presents a new approach to recognize human postures in video sequences comparing two methods. We first describe these methods based on 2D appearances. The one uses projections of moving pixels the reference axis. second method decomposes silhouette into blocks and learns posture appearances through PCA. Then we use 3D model make previous independent camera position. At end give some preliminary results conclude effectiveness this approach.