Height Measurement as a Session-based Biometric for People Matching Across Disjoint Camera Views

A surveillance session can be defined as the period between an individual’s entrance into a surveillance system, and their exit. Utilising this framework enables personal features, such as clothing and height estimates, to become invariant for the duration of the surveillance session. These can then be used as session-based biometrics for matching an individual uniquely as they move throughout the surveillance system, which may include significantly disjoint cameras. By utilising a hierarchy of biometrics to match individuals one can improve the speed of the overall matching. This eliminates unlikely matches quickly by using biometrics that are fast to calculate first. This paper proposes height estimation as an example of such a session-based biometric that would be used earlier in the hierarchy. It demonstrates how statistical height analysis can be quickly calculated as one of the session-based biometric components used to match the tracks of the same individual through image sequences taken by two disjoint camera views in the same surveillance session.