Profile Face Detection: A Subset Multi-Biometric Approach

This paper presents a detection system which uses information from a profile face detector and an ear detector to better localize true profiles. The system was tested on data from a simulated security checkpoint. Drivers' profiles were detected as their vehicles passed the camera. The detection rate for the multi-biometric detector is 95% with an average of 0.15 false detections per image. On the same data, a conventional detector had a detection rate of 97% with an average of 1.8 false detections per image. It is shown that a multi-biometric detector provides a performance boost over conventional detectors.