Order-Preserving Condensation of Moving Objects in Surveillance Videos

摘要: Vision-based detection of illegal or accidental activities in urban traffic has attracted great interest. Since state-of-the-art online automated algorithms are far from perfect, much research effort on offline video surveillance been made to prevent police security staff observing all recorded frames unnecessarily. To solve the problem, this study focuses condensation, which provides fast monitoring moving objects a long duration videos. Considering computational complexity and condensation ratio as two main criteria for efficient we propose algorithm, consists following: 1) initial by discarding nonmoving objects; 2) intra-GoFM (group with objects) condensation; 3) inter-GoFM condensation. In spatiotemporal static pixels within each GoFM temporal between consecutive GoFMs dropped shorten distances objects. Experimental results show that our saves significant amount loads compared previous methods without sacrificing visual quality.