Authors: Ming-Ching Chang, Jixu Chen, Siwei Lyu, Peter Tu
The purpose of this research is to advance the field of video surveillance as it pertains to a variety of law enforcement needs.
This research focuses on the development of three new video analytics technologies:
1. 3D Video Representation and Event Summarization Front-end: which would allow users to view 3D interactive events at a user-specific angle, with highlights, and in a greater context.
2. One-shot Learning for Action Recognition: which could allow users to recognize behaviors such as gaze directions, expressions, and motion fields of a person.
3. Person Specific Face Recognition: which would further improve recognition accuracy by distinguishing features that are specific to a person of interest, such as visible scars and hair styles.
The findings suggest that 3D video analytics technologies are effective and could advance video surveillance in law enforcement by providing a new form of recognition at a glance, allowing for more complex multi-camera imagery.
The study found that this new form of event recognition, using analytics engines, will be able to detect new types of behaviors with as few as a single example. In addition, by considering person specific cues such as hair styles and facial markings, it will become possible to more accurately detect specific persons of interest.
The research implications of this work indicate that surveillance systems of the future will be able to keep pace with the ever-evolving demands on law enforcement.