Acurate segmentation of optic disc is the key to the computer aided diagnosis of glaucoma.Retinal fundus images have large individual differences,complex structures,and in which optic disc is easily interfered by vascular tissue and pathological changes,so the accurate segmentation of optic disc is difficult.This paper proposes an optic disk segmentation method based on superpixel classification and dynamicoptimization of radial ray scanning.The super pixels are classified by using gray scale,texture,anatomical structure and other features.Radial ray projection scanning is used to dynamically optimize the precise position of the disc contour.The method of optic disc segmentation based on super pixel classification has strong inhibition to noise interference such as blood vessels and lesions,can realize rough segmentation of optic disc.The dynamic optimization method of radial ray projection scanning makes full use of the geometric structure,edge,curvature,continuity and other characteristics of the optic disc to ensure the integrity and accuracy of optic disc region segmentation.The experimental results show that REFUGE database and DRIONS database,the accuracy is 99.19%and 98.65%respectively,the sensitivity is 98.64%and 94.99%respectively,the specificity is 99.46%and 99.89%respectively,the DICE coefficient is 98.52%and 97.28%respectively,and the area overlap is 97.09%and 94.70%respectively,compared with typical algorithms,we proposed method has significantly improved the accuracy,sensitivity and specificity of optic disk segmentation on different databases.