Research On Abnormality Detection And Abnormality Synopsis In Surveillance Videos

In this paper, we propose a new approach to detect abnormal activities in surveillance videos and creat e suitable summary videos acco rdingly. The proposed approach first introduces a patch-based method to automatically model normal activity pattern s and key regions in a scene. In this way, abnormal activities can be ef fectively detected and classified from the modeled normal patterns and key regions. Then, a blob sequence optimization process is proposed which integrates spatial, temporal, size, and motion correlation among objects to extract suitable foreground bl ob sequences for abnormal objects. W ith this process, blob extraction errors due to occlusion or background interfe rence can be ef fectively avoided. Finally, we also propose an abnorm ality-type-based method which creates short-period summary videos from long-period input surveillance videos by properly arranging abnorm al blob seque nces according t o their activity types. Experimental results show that our proposed approach can effectively create satisfying summary videos from input surveillance videos.