Automated Person Categorization in Video Using Soft Biometrics

In this proposal, intuVision and partner West Virginia University Biometric Research Group propose to research, design, and develop a feasibility prototype for a Video Tracking and Object Re-Identification (VTORI) system that can support entity re-identification among multiple sensors and databases. Firstly, we will focus on surveillance video data, explore soft biometry features that can be robustly extracted from different types of video sources and reliably used for subsequent analysis and implement these soft biometry features into our team’s existing video product test platforms to validate our concept prototype. Secondly, we will create a metadata dictionary and a standard description scheme for soft biometry features to facilitate efficient exchange of extracted identification information between different sensors and systems. Our approach is built on our team’s expertise and leverages our previous and current related work in video object tracking, feature extraction for video object classification and camera hand-off of tracked objects and study of face, gait and human metrology based biometric features from video.