Real-Time Face Recognition from Surveillance Video - Intelligent Video Event Analysis and Understanding - page 194

same person. (This feature was not implemented in our system). The feature vectors
are then packaged for transmission to the back-end using TCP/IP.
In section 3.3, we described how the front-end should attempt to match faces to
both a frontal and a profile model of the individual. In the design of our system,
each match would be performed by an independent thread, as shown in figure 32.
The “second thread” in the front-end is in fact two separate threads, one detect-
ing left-facing profiles and one detecting right-facing profiles. Thus, three threads
would search for frontal faces, right profiles and left profiles, respectively. A fourth
thread handles all the communication with the back-end. (Profile detection was not
implemented in our system; see section 3.3.5 for discussion).
Fig. 32 Front End: Multi-threaded approach to feature extraction and matching against
frontal face model and profile face model