Key Technologies For Pupil Location On Eye Area Image

Pupil location technology of eye region has been one of the important research areas in computer vision and pattern recognition. It has been widely used in perimetry, eye tracking, virtual reality, human-computer interaction, biometrics, intelligent transportation fields. With the continuous progress of science and technology and the rapid application of computer vision technology in commercial fields, In the case of subjects without interference, access to the position and radius of pupil in the eye area image accureately and rapidly has attracted more and more researchers concerned. In this paper, under the background of pupil location in perimetry, proposes an efficient method of pupil position in eye area image. Main work and innovation of this paper is as follows:(1) Propose a novel thresholding method for pupil in eye area image. This method uses the characteristics of histogram curve, combined with Gaussian derivatives, automatically finds out the location of the first significant valley point belong to the histogram curve, and corresponding gray value of the valley taken as the threshold, so as to effectively extract the pupil region. At the same time, this algorithm can also be used as a multi-threshold segmentation algorithm applied to other areas.(2) Compare several gradient calculation method, Analysis of the relationship between gradient computation method based on gaussian derivative and traditional Sobel and Prewitt gradient operator. Prove the Sobel operator is a special example of gradient computation method based on gaussian derivative nuclear in nature and summarize the choice of these methods.(3) The edge of pupil area obtained from image is corrected by the gradient direction. To get more real edge pixels of the pupil, moving the edge pixels has a smaller gradient magnitude to the position of edge pixels has a greater gradient magnitude.(4) Propose a randomized circle detection algorithm constrained by gradient direction and demonstrate the effectiveness of the algorithm by experiments. As a result, this method requires less time and has strong noise immunity when compared with other circular detection algorithm.(5) Presents a complete set of pupil location algorithm on the eye area image. Firstly, get the binary image of pupil area through threshold segmentation method based on the characteristics histogram that proposed in this paper. Then obtain the edge collection of binary image by edge tracking and correct the position of each pixel in the edge set with the gradient information. Finally locate pupil position using the gradient constrained randomized circle detection algorithm.