Title :
Annotation of fighting scenes with moving and changing backgrounds
Author_Institution :
Sch. of Comput. Sci. & Eng., Univ. of Electron. Sci. & Technol. of China, Chengdu, China
Abstract :
In this paper, a novel method is proposed to annotate fighting scenes with moving and changing backgrounds. The proposed method is implemented in four major steps. Firstly, points of interests are extracted and traced to build the trajectory framework. Secondly, the global motion is estimated using trajectory records. Thirdly, camera motions are detected and classified. Fourthly, the behaviours of the points of interest are labelled by fuzzy scores and a voting scheme is proposed to annotate fighting scenes. Extensive experiments were carried out using PETS´04 database and Hollywood action films. The proposed method achieves satisfactory performances.
Keywords :
cameras; fuzzy set theory; image sequences; motion estimation; video signal processing; Hollywood action films; PETS´04 database; camera motions; changing backgrounds; fighting scenes annotation; fuzzy scores; global motion estimation; moving backgrounds; trajectory framework; trajectory records; video sequences; voting scheme; Cameras; Conferences; Films; Hidden Markov models; Humans; Tracking; Trajectory; Points of interests; fighting; motion; trajectory;
Conference_Titel :
Apperceiving Computing and Intelligence Analysis (ICACIA), 2010 International Conference on
Conference_Location :
Chengdu
Print_ISBN :
978-1-4244-8025-8
DOI :
10.1109/ICACIA.2010.5709899