Real-time iris tracking with a smart camera
Mehrübe Mehrübeoğlu, Ha Thi Bui, Lifford McLauchlan · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2011
This paper presents a real-time iris detection procedure for gray intensity images. Typical applications for iris detection utilize template and feature based methods. These methods are generally time and memory intensive and not applicable for all practical real-time embedded realizations. Here, we propose a method that utilizes a simple algorithm that is time-efficient with high detection and low error rates that is implemented in a smart camera. The system used for this research involves a National Instruments smart camera with LabVIEW Real-Time Module. First, the images are analyzed to determine the region of interest (face). The iris location is determined by applying a convolution-based algorithm on the edge image and then using the Hough Transform. The edge-based less complex and less computationally expensive algorithm results in an efficient analysis method. The extracted iris location information is stored in the camera's image buffer, and used to model one specific eye pattern. The location of the iris thus determined is used as a reference to reduce the search region for the iris in the subsequent images. The iris detection algorithm has been applied at different frame rates. The results demonstrate the speed of this algorithm allows the tracking of the iris when the eyes or the subject is moving in front of the camera at reasonable speeds and with limited occlusions.