The invention discloses an
ellipse-detection-based
iris detection method. The method comprises: step one, obtaining an image including an eye region; and step two, carrying out
ellipse detection on the eye region; to be specific, step 2-1, carrying out
ellipse detection on the eye region by using an ELSD
algorithm; to be specific, changing the sensitivity of the ELSD for N times to obtain information of N groups of circles; step 2-2, according to the obtained circle information, constructing a circumference by using a
point set of circumferential points in the circle information; step 2-3, for each point of the
point set, constructing feature vectors of the point and selecting feature parameters, including a circle center coordinate, a
radius, an arc degree, an
arc length, and a circumferential point coordinate, of the point to carry out weighted
processing to obtain feature vectors; step 2-4, classifying the feature vectors, using one class with the largest individual number as a target class, and storing circumferential point coordinates of all feature vectors; and step 2-5, carrying out
circle fitting by using the circumferential point coordinates in the step 2-4.