基于激光点云的阔叶树叶片重建与形变研究

发布时间：2019-01-17 19:34

【摘要】：从三维点云数据中重建出高精度的植物叶片模型并进行叶片形变模拟一直是林学、计算机图形学、植物生理生态学、景观设计等众多学科研究的热点问题之一。植物叶片不仅是植物最重要的器官之一,同时也是植物叶片生理功能、形态特征分析、冠层光分布计算等研究的基础。但由于室外植物叶片的点云数据存在噪声、叶片边缘不光顺、叶片表面存在孔洞等问题,所以不易重建出高精度的植物叶片模型。针对存在的问题,本文首先对室外植物叶片点云数据进行去噪处理,然后根据植物叶片点云数据的边缘点云数据拟合出新的叶片边缘点云数据,其次对叶片表面进行拟合并采用Delaunay三角剖分算法生成植物叶片的网格模型,最后对存在孔洞的植物叶片网格模型进行修补。另一方面围绕植物叶片的三维形变模拟研究已经有了几十年的历史,经过几十年的研究努力,目前已经取得了显著的成果。但在这些研究中,大部分植物叶片模型都是通过虚拟现实技术生成的,植物叶片模型也就不具有真实性。且植物叶片模型形变往往只是采用简单的弹簧-质点模型实现叶片形变,方法过于简单、单调。针对以上问题,本文首先将植物叶片网格模型实体化并添加主脉,然后对叶肉部分和叶脉部分赋予不同的材料属性,最后采用目前比较流行的几种形变方法,实现植物叶片模型形变。本文主要研究内容如下:(1)点云数据去噪方法研究。提出了一种适合于室外植物叶片点云数据的去噪算法,该算法将噪声点分为三类。第一类是距离大片中心点云较远、小而密集的点云。第二类是偏离点云较远,悬浮在点云上方的离散、稀疏点。第三类是噪声点和真实点混合在一起的点,该类点云的形态呈分层状。并分别对这三类噪声点进行去噪。(2)植物叶片边缘拟合算法研究。提出了一种适合于室外植物叶片点云数据的边缘拟合算法。该算法首先提取出植物叶片的边缘点云数据,然后根据空间曲线的定义拟合出光顺的植物叶片边缘。(3)植物叶片表面拟合与重建方法研究。提出了一种适合于室外植物叶片点云数据的表面拟合与重建方法。该方法首先采用双三次Bezier曲面拟合叶面,使叶片表面更加光顺,然后通过二维Delaunay算法实现植物叶片的三维重建。(4)植物叶片网格模型表面孔洞修补方法研究。提出了一种适合于室外植物叶片点云数据的表面孔洞修补方法。该方法首先识别出植物叶片网格模型中的孔洞边缘,然后通过波前法实现孔洞的修补。(5)植物叶片形变方法研究。提出了一种适合于室外植物叶片点云数据的叶片实体化方法,并将目前流行的几种形变方法运用到植物叶片模型中,以实现室外植物叶片模型形变方法的多样性。
[Abstract]:It has been one of the hot issues in forestry, computer graphics, plant physiology and ecology, landscape design and so on to reconstruct high precision plant leaf model from 3D point cloud data and simulate leaf deformation. Plant leaves are not only one of the most important organs of plants, but also the basis of physiological function, morphological characteristics analysis and calculation of canopy light distribution. However, due to the noise in the point cloud data of outdoor plant leaves, the edge of the leaves is not only smooth, there are holes in the surface of the leaves, so it is difficult to reconstruct the high precision plant leaf model. In order to solve the existing problems, the data of outdoor plant leaf point cloud are de-noised firstly, and then the new data of leaf edge point cloud are fitted according to the edge point cloud data of plant leaf point cloud data. Secondly, the meshes of plant leaves were generated by Delaunay triangulation algorithm, and the mesh models with holes were repaired. On the other hand, the three-dimensional deformation simulation research around plant leaves has been decades of history, after decades of research efforts, has achieved remarkable results. However, in these studies, most plant leaf models are generated by virtual reality technology, and the plant leaf models do not have authenticity. And the deformation of plant leaf model is usually realized by simple spring-mass model. The method is too simple and monotonous. In order to solve the above problems, the model of plant leaf grid is first materialized and the main vein is added, then the mesophyll part and the vein part are given different material properties. Finally, several deformation methods which are popular at present are adopted. The deformation of plant leaf model was realized. The main contents of this paper are as follows: (1) the method of point cloud data denoising. A denoising algorithm for outdoor plant leaf point cloud data is proposed, which divides the noise points into three categories. The first is a small, dense point cloud that is far from the center of a large area. The second is the discrete and sparse point suspended above the point cloud, which is far away from the point cloud. The third is the point where the noise point and the real point are mixed together, the shape of this kind of point cloud is stratified. The three kinds of noise points were de-noised respectively. (2) the edge fitting algorithm of plant leaves was studied. An edge fitting algorithm for outdoor plant leaf point cloud data is proposed. Firstly, the edge cloud data of plant leaves are extracted, and then the smooth leaf edges are fitted according to the definition of spatial curve. (3) the method of surface fitting and reconstruction of plant leaves is studied. A surface fitting and reconstruction method for outdoor plant leaf point cloud data is proposed. In this method, the surface of the leaf is fitted with a bicubic Bezier surface to make the leaf surface more smooth, and then the three-dimensional reconstruction of the plant leaves is realized by using the two-dimensional Delaunay algorithm. (4) the method of repairing the holes in the surface of the plant leaf mesh model is studied. A surface hole repair method for outdoor plant leaf point cloud data is proposed. In this method, the hole edges in the plant leaf mesh model are first identified, and then the holes are repaired by the wavefront method. (5) the deformation method of plant leaves is studied. A method of leaf materialization suitable for outdoor plant leaf point cloud data is proposed, and several deformation methods are applied to plant leaf model in order to realize the diversity of outdoor plant leaf model deformation methods.
【学位授予单位】：南京林业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41

【参考文献】