Video Summarization via Cluster-Based Object Tracking and Type-Based Synopsis

Yuxi Li, Weiyao Lin, Tao Wang, Qi Guo, Ruijia Yang, Shugong Xu

Abstract: In this paper, we construct a trajectory-based system for the synopsis of surveillance video stream. The proposed approach first applies a cluster-based tracking method to extract foreground object from input videos, then extracts the abnormal object and classifies them into different motion patterns. Finally a type-based synopsis scheme is proposed to properly gather the moving object of different pattern types into limited time endurance. As a consequence, this system would be helpful for accurate and fast surveillance videos analysis.