With the development of technologies such as machine vision, depth cameras are becoming more and more widely used in 3D modeling. For the phenotypic surface reconstruction of plants, this paper designed a method to obtain the point cloud data of the plant phenotype by Kinect, and collected images at a certain angle interval and stitched into a complete point cloud image by rotating the point cloud coordinates.Then used straight-through filtering and statistical filtering to complete the filtering of the plant point cloud image.Then used the greedy projection triangulation method to reconstruct the surface of the plant point cloud. Finally, the 3D reproduction of the obtained surface model was implement with the help of MeshLab software as well as the measurement of blade length parameters. Experimental results show that there is no significant difference between the phenotypic parameters of the reconstructed plant surface model and the measured values, and the relative error is kept within 5%, this method can meet the needs of 3D reconstruction of plant phenotype.