Research On Face Detection And Pupil Center Location

At present, eye detection has become one of the research hotspots in the field of computer vision, widely applied in eye tracking, human computer interaction, disease diagnosis, industrial engineering and human for analysis, psychology, physiology. Currently, the application of gaze tracking technology is stricter to the accuracy of the human eye location.The eye detection and pupil center location is based on face detection. This paper combines integral projection and radial symmetry transform algorithm to locate human eyes from coarse to fine. The paper improves the traditional face and eye detection method, and studies the relevant problems of the face and eye detection and localization. The main work of this paper is as follows: 1. Face detection and localizationAn ellipse skin color model based on adaptive light threshold is proposed, which is based on the analysis of the traditional elliptical skin color model in the YCbCr color space. Firstly, the face image is converted into YCbCr color space from RGB color space; secondly, Skin color segmentation is carried out based on an ellipse skin color model based on adaptive light threshold; finally, morphological open operation and area filtering method is used to filter the pseudo color region in the background and face area is marked with rectangular box. Experiments show that the improved skin color model proposed in this paper improves the effect of skin color segmentation.2. Human eye detection and locationConsidering the questions that the projection curve varies irregularly because of noise interference when using horizontal and vertical integral projection directly to detect human eyes, it is difficult to locate eyes precisely from projection curve directly. A joint of integral projection and radial symmetry transform is proposed to locate the human eye from coarse to fine. Firstly, the horizontal integral projection is used to detect the candidate region of the eye in the detected face region, which is the rough positioning of the human eye. Then, locate the pupil center by the radial symmetry transform in the left and right eye regions detected from face, which is the precise positioning of the human eye. Experiments show that compared with the traditional integral projection algorithm, the proposed algorithm improves the accuracy of eye location.