A fast video forensics method based on face detection

作者：Zhou Zhiping; Yang Haibin^*

来源：Journal of Information and Computational Science, 2013, 10(1): 79-86.

• Background subtraction
• Color features
• Face detection algorithm
• Face images
• Fast video
• Gaussian Mixture Model
• Human bodies
• Human faces
• Interference factor
• Object Tracking
• Point problem
• Real time requirement
• Surveillance video
• Video forensics

This paper puts forward a fast and effective method to detect particular face in surveillance video for video forensics. At first, in order to solve the starting point problem of surveillance video, the first frame is abstracted to detect face using the face detection algorithm based on complexion model and eye feature. And then, the Gaussian Mixture Model for background subtraction is used to extract the region of moving human body so that to reduce the interference factors of similar complexion when detecting face. The cache queue based on color feature for objects matching is set up to replace object tracking and that greatly reduces the forensic time. Finally the SURF algorithm and Gabor filter are used to extract features from face image for matching the query human face with detected human face so that to locate the position that the query face appears. Experiment results express that the particular face detection in video based on SURF algorithm and Gabor filter is fast and efficient. So this method can meet the real-time requirement and can provide further evidences for video forensics.

• 出版日期2013
• 单位江南大学