The ratio and the diameter of artery/vein of the retinal can reflect the risk of stroke of hypertension, thus, quantitative analysis of the diameter of the vessel can contribute to risk assessment and prevention of the disease. A artery/vein automatic classification method in Retinal Images and Vessel Diameter Measurement are introduced in this paper. Firstly, the vessel centerline is obtained after segmenting the retinal vessel. Secondly, different channel components in different color spaces are selected, the eigenvectors of centerline pixels, vessel pixels. vessel widths and central light reflection are defined. Kmeans clustering algorithm is adopted to classify the artery/vein in the measurement region of interest. Finally, the distribution curve of the vessel cross section is fitted using the Gaussian curve and the width of the artery/vein can be obtained according to Half Height Full Width. Experiments with Review public database and Drive public database prove the effectiveness of the proposed method.