摘要
This paper describes a robust system for human face recognition designed by the authors. In this realistic real time criminal recognition system, used at customs, airports and harbors, features extracted from face images are individual, hidden, light invariant, and effective to recognition. Features based on KL transform instead of features on spatial domain, such as template, shape, outline, height, width, area, overcome the sensitivity of illumination for inputting images, and improve the rate and speed of recognition. A set of preprocessing like locating before feature extraction normalize images as well.
０ ＩＮＴＲＯＤＵＣＴＩＯＮＨｕｍａｎｆａｃｅｒｅｃｏｇｎｉｔｉｏｎｉｓａｎｉｍｐｏｒｔａｎｔｔｏｐｉｃｆｏｒｃｏｍｐｕｔｅｒｖｉｓｉｏｎｒｅｓｅａｒｃｈａｎｄａｌｓｏｉｓｏｎｅｏｆｔｈｅｍｏｓｔｄｉｆｆｉｃｕｌｔｐｒｏｂｌｅｍｓｉｎｔｈｅｆｉｅｌ...
