Abstract
Security personnel are interested in detecting persons who are loitering in a region captured by a security camera. However, in many scenarios, it is possible that a person walks outside the region captured by a security camera for a period of time before returning to the region. The intermittent presence of the person may not be detected as an instance of loitering. This disclosure describes techniques to detect intermittent loitering in security camera footage. Per the techniques, the period for which a person leaves the screen, if below a certain threshold, is counted towards loitering time. The techniques are face-agnostic, such that loitering is detected without the need for face recognition and are highly scalable. Face-agnostic embeddings of persons detected in the camera footage are obtained and clustered. A person appearing for a total time exceeding a first configurable threshold is declared to be loitering if gaps in their appearance are less than a second configurable threshold.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
NA, "Face-agnostic Intermittent Loitering Detection in Security Camera Footage", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/9514