Referring to the complexity of the eyelid edge structure, many interferential factors and the slowness of the localization speed, large window smoothing filtering, edge points detection in special directions and high threshold edge point screening were used to reduce the interference from the eyelash, eye texture and other ridge-like edge, and the point set was acquired on the edge of the single-layer or the double-layer eyelid rapidly. And then, the curve fitting method was used to process the detected point set to acquire the continuous eyelid edge curve. The first eyelid location was completed. Finally, according to the characters that the inner and outer edge of double-layer eyelid are approximate parallel and the gray distribution changes sharply on the edge of inner eyelid, take the fitting curve acquired in first eyelid location as a template, and shift it down parallel to detect the position where gray value changes abruptly. Then the second time eyelid location was finished. The experiment shows that the location speed can be improved effectively on the basis of ensuring an high location accurate rate.