The Broad-leaved Tree Leaves Reconstruction And Deformation Based On Laser-point-clouds

To reconstruct high-precision models of plant leaves and to simulate leaf deformation from 3D point cloud data is one of the hot issues in many subjects such as forestry, computer graphics, plant physiological ecology, and landscape design. Leaves are not only one of the most important organs of the plant, but also the research foundation of plant leaf physiological function, morphology analysis, and computation of canopy light distribution. However, it is not easy to reconstruct the precision model of plant leaves because of the noise in the point cloud data, the roughness of the leaf edges, and the holes in leaf surface.In this article, aimed at the existing problems, the point cloud data of the outdoor plant leaves are denoised first. Then, the new point cloud data of leaf edges are fitted according to the original leaf edges. After that, the leaf surface is fitted and the grid models of plant leaves are generated using Delaunay triangular subdivision algorithm. Finally, the grid models of plant leaves which have holes are repaired.On the other hand, 3D deformation simulation research around plant leaves already has decades of history and after decades of research efforts, significant progress has been made. But in these studies, most models of plant leaves are generated by virtual reality technology, so the models do not have authenticity. Moreover, the deformation of plant leaves are often realized using the spring-particle model, which is too simple and monotonous.To solve above problems, the grid models of plant leaves are materialized and added with the main vein first in this article. Then, the mesophyll parts and vein parts are given different material properties. Finally, the deformation of the plant leaves is realized using several deformation methods which are relatively popular at present.The main research contents in this thesis are as follows:(1) Research on point cloud data denoising methods. We put forward an algorithm suitable for denoising point cloud data of outdoor plant leaves, which divide noises into three categories. The first is small and dense point cloud, far from the center of large cloud. The second category is discrete, sparse point deviating far from the point cloud, and suspension on top of the point cloud. The third kind is the noise points mixed with real points, and the point cloud is stratified. And the three kinds of noises are denoised respectively.(2) Research on the plant leaf edge fitting methods. We put forward a kind of method suitable for fitting point cloud data of outdoor plant leaves. In the algorithm, the edge point cloud data of the plant leaves are extracted firstly, and then the edge of smoothing plant leaves are fitted according to the definition of the space curve.(3) Research on plant leaf surface fitting and reconstruction methods. We present a kind of algorithm suitable for surface fitting and reconstructing the point cloud data of outdoor plant leaves. In this method, the leaves are fitted by adopting bicubic Bezier curve surface, making the leaf surface smoother, and then 3D reconstruction of plant leaves is realized by 2D Delaunay algorithm.(4) Research on repairing holes in grid models of plant leaf surface. We put forward a method suitable for repairing holes in the surface of point cloud data of outdoor plant leaves. This method firstly identify the hole edges in the grid model of plant leaves, and t hen achieve the hole reparation through the wave front method.(5) Research on the deformation methods of plant leaves. We present a suitable algorithm for materialization of point cloud data of outdoor plant leaves, and several current epidemic deformation methods are applied to the model of plant leaves to realize the diversity of deformation methods for models of outdoor plant leaves.