Generalised Spatio Temporal Feature Based Important Activity Synopsis Generation

摘要: Traditional video analysis methods can generate summary of day's long videos. While generating synopsis maintaining the motion structure important activities present in a sequence is great concern research communities and industry. In this paper, we an automatic scalable approach for detection based on different spatiotemporal criteria used as signature. To maintain context cues propose online preserved approach, which retain behavior interactions between objects original while condensing much content possible. A hierarchical fashion employed to efficiently search sequence, those both spatial collision temporal consistency are considered. Experimental results numerous (six) sequences demonstrate promise proposed approach.