Proliferative diabetic retinopathy is the more serious conditions as it involves the proliferative growth of abnormal new vessels on the retina. New vessels characteristic appearance like narrower calibre and more tortuous. This paper presents a method for detecting new vessels on the optic disc based on the SVM classifier. First, vessels like candidate segments are extracted using green component, because in the colour retinal images the blood vessels appear most contrasted in the green channel compared to red and blue channels in RGB image. Second, candidate new vessel segments re detected using a method of morphological watershed transform. Third, fifteen feature parameter are calculated for each segment. Based on these feature, each segment is classified as normal or abnormal vessels using SVM classifier. The system was trained with normal and abnormal retinal images.