Eye detection is very important for automatic face recognition and gaze tracking. In this paper we propose an algorithm for eye detection under active infrared (IR) illumination, which will be integrated into our facial analysis system for cockpit access control. Based on a sophisticated thresholding algorithm, pupil candidates can be effectively selected. Formerly we used circle matching to verify pupil candidates, which has difficulty in distinguishing pupils from the interior corner of the eye and some background noise. Seeing that in this paper we utilize an appearance model, which is created by using Principal Component Analysis (PCA) and Support Vector Machines (SVM). This appearance model, trained using PCA and SVM, is exploited to verify the pupil candidates, which performs better than our previous approach of circle matching. However, the existence of eyeglasses has a negative effect on candidate selection. There are also cases where pupils are not bright enough. Regarding these problems the Radial Symmetry Transform (RST) is exploited. To increase the robustness against eyeglasses and pupil intensity, the radial symmetry transform is incorporated, which performs more efficiently and more accurately than the generalized symmetry transform used in our previous work. The experimental results demonstrate the effectiveness of the presented eye detection approach.
Robust eye detection for cockpit access control using an appearance model and radial symmetry transform under active near infrared illumination
Robustes Verfahren zur Augendetektion für die Zugangskontrolle im Cockpit durch Verwendung eines Modells des Erscheinungsbildes und der Transformation der Radialsymmetrie unter Nah-Infrarotbeleuchtung (NIR)
2007
11 Seiten, 9 Bilder, 3 Tabellen, 11 Quellen
Aufsatz (Konferenz)
Englisch