The accurate classification of plant organs is a key step in monitoring the growing status and physiology of plants. A classification method was proposed to classify the leaves and stems of potted plants automatically based on the point cloud data of the plants, which is a nondestructive acquisition. The leaf point training samples were automatically extracted by using the three-dimensional convex hull algorithm, while stem point training samples were extracted by using the point density of a two-dimensional projection. The two training sets were used to classify all the points into leaf points and stem points by utilizing the support vector machine (SVM) algorithm. The proposed method was tested by using the point cloud data of three potted plants and compared with two other methods, which showed that the proposed method can classify leaf and stem points accurately and efficiently.