Eye detection based on rank order filter and projection function

Abstract

Eye detection is the preconditions for fatigue driving detection. In this paper, considering the deficiency of the origin algorithm based on rank order filter, an improved eye detection algorithm based on rank order filter and projection function is proposed. At first, according to the structure features and gray features, an ellipse mask combined with rank order filter is used to find the eye candidates initially. Then projection function is applied to excluding the no eye candidates and exactly locating eyes. By the comparative experiment of eye detection in the BioID face database, the results have shown the higher accuracy and efficiency of the improved algorithm.