This paper describes a method f o r quickly and robustly localizing the iris and pupil boundaries of a human eye in close-up images. Such an algorithm can be critical for iris identifcotion, or for applications that must detemzine the subject's gaze direction, e.g., human-computer interaction or driver attentiveness determination. A multi-resolution coarse-toTfne search appmach is used, seeking to maximize gradient strengths and uniformities measured across rays radiating from a candidate iris or ptipil's cenrral poinr. An empirical evaluation of 670 eye images. both with and without glasses, resulted in a 98% localization accuracy. The algorithm has also shown robustness tu weak illuminarion and mosf specular repertions (e.g., at exewear and cornea), simplij+)ing system component requirements. Rapid execution is achieved on a 750 M H z desktop processor

lgorithms based on the correlation of image patches can be robust in practice but are computationally intensive due to the computational complexity of their search-based nature. PerformingAt he search over time instead of over space is linear in nature, rather than quadratic, and results in a very efficient algorithm. This, combined with implementations which are highly efficient on standard computing hardware, yields performance of 9 frames/sec on a scientific workstation. Although the resulting velocities are quantized with resulting quantization error, they have been shown to be sufficiently accurate for many robotic vision tasks such as time-to-collision and robotic navigation. Thus, this algorithm is highly suitable for real-time robotic vision research.

A real-time robot vision system is described which uses only the divergence of the optical $ow field for both steering control and collision detection. The robot has wandered about the lab at 20 c d s for as long as 26 minutes without collision. The entire system is implemented on a single ordinary UNIX workstation without the benefit of real-time operating system support. Dense opticaljlow data are calculated in real-time across the entire wide-angle image. The divergence of this optical $ow field is calculated everywhere and used to control steering and collision behaviol: Divergence alone has proven sufJicient for steering past objects and detecting imminent collision. The m j o r contribution is the demonstration of a simple, robust, minimal system that uses $ow-derived measures to control steering and speed to avoid collision in real time for extended periods. Such a system can be embedded in a general, multi-level perceptiodcontrol system.

No abstract

No abstract