A method for 3D reconstruction of apple tree LB based on point cloud data

Abstract.

Three-dimensional (3D) reconstruction of leaf blade (LB)morphological structure is important for the study of morphological characteristics of LB and calculation of tree canopy light distribution. In this paper, a 3D reconstruction method of apple tree LB has been proposed based on the point cloud data. First of all, an appropriate 3D scanner was chosen to collect the point cloud data according to the morphological characteristics of the leaves. Then the noises of the original point cloud data was removed utilizing manual method while the point cloud simplification was completed by curvature-based point cloud simplification (CBPCS) method. The CBPCS method consists of four stages The CBPCS consists of four stages. The first three stages are the establishment and division of bounding boxes, k neighborhood search and curvature calculation (including Mean curvature, Gaussian curvature and global mean curvature). The last stage is point cloud simplification in accordance with the prescribed principle. Compared with the global mean curvature, if the local mean curvature is less than the global mean curvature, then r points are randomly reserved (r is adjusted according to the reduction rate). If the local mean curvature is greater than the global mean curvature, all points with curvature greater than the local average curvature are preserved. Finally, 3D surface reconstruction was completed. The leaf area of the reconstructed model was compared with that of the real leaf, and the accuracy is higher than 99%. The results show that the reconstruction model could better maintain the characteristics of the leaves and provide some references for the 3D reconstruction and the visualization of the canopy.