Video Frame Classification Based upon Motion Vectors
Ting-Yu Hsu；Wen-Kuei Lin；Che-Yen Wen
Surveillance Systems ； Video Frame Classification ； Crime Investigation ； Motion Vectors ； Gaussian Mixture Model
Forensic Science Journal
|Volume or Term/Year and Month of Publication||
14卷1期（2015 / 12 / 01）
1 - 32
Closed-Circuit Television (CCTV) surveillance systems play an important role in crime scene investigation. Investigators use recorded video data to find suspects and information to solve crime cases. However, there are usually a lot of unwanted or tedious frames (e.g., background without meaning) in surveillance videos. It is an exhausting and time-consuming work for investigators to search entire videos for interested information. The most difficult problem of video data is how to find useful and meaningful frames from them. In this paper, we provide a method of utilizing motion vectors to classify video frames based upon moving information (such as orientation). By the proposed method, investigators can reduce the time of searching objects from surveillance videos, and focus on frames with moving objects.