Fast Pupil center localization system based on SSD Cascade gradient

In order to solve the problem of low recognition and high misrecognition of traditional eye tracking system, which uses cascade classifier to obtain face image. This paper proposes a model of SSD(Single Shot MultiBox Detector) combined gradient algorithm. The method, firstly, put the SSD in depth study of facial model to replace the cascade classifier, with a face image segmentation the eye part of the detected image. Secondly using the improved gradient localization algorithm to locate the pupil center position, and then through the proposed simple judgment mechanism on the rationality of the pupil center again decrease the misrecognition, finally get the pupil center. Experimental results show that the proposed algorithm can achieve a detection rate of 5.42 frames per second and improve the detection accuracy by 6%.