CASIA OpenIR  > 毕业生  > 博士学位论文
Alternative TitleFast plant modeling based on image and sketch
Thesis Advisor张晓鹏 ; 胡包钢
Degree Grantor中国科学院研究生院
Place of Conferral中国科学院自动化研究所
Degree Discipline计算机应用技术
Keyword植物建模 图像 草图 快速建模 三维骨架建模 Plant Modeling Image Sketch Fast Modeling 3d Skeleton Modeling
Abstract植物是自然场景的重要组成部分,植物建模是虚拟现实领域的研究热点和难点。植物建模是以自然界的植物为模型,以激光扫描、相机拍摄和手工交互等为获取信息的手段,用图形学方法创建出植物的三维几何形态的过程。由于植物种类多样、结构复杂,现有的方法一般建模时间长,普通用户不易使用,不能满足计算机游戏、三维电影等应用领域对视觉效果和建模速度的双重要求。本文根据实际应用需求提出了几种基于图像和草图的简单快速的植物建模方法。本文的主要工作和贡献有: 1.提出了两种基于透视投影粒子流的植物建模方法——基于轨迹分析和透视投影粒子流的方法、基于叶尖点和透视投影粒子流的方法。这两种方法由图像配准信息合成三维矢量场,通过矢量场中的粒子运动实现植物枝干形状的模拟。这些技术的创新性在于将视觉重建与粒子流模拟相结合,用于枝干末梢位置信息提取和粒子运动轨迹约束,从而使重建的三维枝干的位置和形状与输入图像信息相吻合。 2.提出了一种基于深度检索的树木建模方法。该方法以用户勾画的草图作为输入,通过深度检索构建三维主枝,基于自相似原理在二维树冠轮廓约束内产生各级细枝。这种方法的技术创新在于首次提出了深度检索方法。深度检索方法解决了两个输入草图中二维枝干的对应问题,使重建的三维枝干能够严格保持两个二维输入的形状信息。该方法避免了现有方法适用树木形态有限、不能反映树冠中树枝形状及重建的枝干与输入图片有明显差异等问题。 3.提出了一种基于骨架点云的树木建模方法。该方法基于等高同位置对应原则从两幅草图构建反映树木三维形态的骨架点云,通过匹配点搜索将一个二维骨架扩展为满足点云形态约束的三维骨架。此方法的技术创新在于新的骨架点云构建方法解决了三维信息恢复中的枝干遮挡与对应匹配问题,与现有方法相比,提高了建模速度。 4.提出了一种基于密度控制的树木建模方法。该方法通过在多个高度平面控制枝的分布密度将一个二维骨架转化为一个三维骨架。该方法的创新性在于新的深度密度控制方法可以实现树枝在深度方向的分布和形状控制,避免了单幅图像建模的夹角最大方法只能产生对称形状树木模型的问题。 5.研究并实现了一个树木建模系统平台。该平台以草图作为输入,分别用基于深度检索的方法和基于骨架点云的方法构建三维树木模型。它使用方便,建模效率高。作为一种快速的树木建模工具,该平台能够满足虚拟现实应用对模型效果和建模速度的要求,并为新的植物建模方法的研究提供了运行环境。
Other AbstractSince plants constitute an important part of natural scene, plant modeling is widely studied in the field of virtual reality. Plant modeling is a process that shape information is gathered through scanner, camera, or interaction, and plant structure is constructed using methods in computer graphics. Due to structural complexity and botanical diversity, current methods for plant modeling are generally time-consuming and difficult to use for novices, and they can’t satisfy the actual applications in computer games and 3D films, in the aspects of visual effect and speed. In this paper, we present several simple methods for fast plant modeling according to application needs. The main work and contributions are as follows: 1. Two plant modeling methods based on perspective particle flows are put forward. One method is based on trace analysis and perspective particle flows and the other method is based on leaf apex and perspective particle flows. In each method, a 3D vector field is constructed from image registration and a plant skeleton is simulated through particle flows. The innovative idea of these techniques is that they combine visual reconstruction with particle flows, and use them for particles’ initial position computation and trace restriction. The shape and position of 3D branches are accordant with their input images. 2. A tree modeling method using depth retrieval is proposed. From sketches of users, main branches are constructed using depth retrieval, and within 2D crown silhouette, small branches are modeled based on the principle of self-similarity. A depth retrieval method is put forward. It resolves the problem of branch matching in two sketches, and constructs a 3D structure that keeps the shapes of both input sketches accurately. Some problems in current methods, such as narrow application range, lack of shape control for branches in leaves, and difference between model and inputs are avoided in our method. 3. A method for tree modeling by building skeleton point cloud is put forward. Based on the matching principle of same height and same position, a 3D skeleton point cloud is built from two sketches. Through searching matching point in the 3D point cloud, a 2D skeleton is converted into a 3D structure which satisfies the restriction of point cloud shape. The advantage of this method is that the problems of branch occlusion and matching are solved, and compared with current methods, modeling efficiency is improved. ...
Other Identifier200618014629078
Document Type学位论文
Recommended Citation
GB/T 7714
刘佳. 基于图像和草图的快速植物建模[D]. 中国科学院自动化研究所. 中国科学院研究生院,2010.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_20061801462907(8375KB) 暂不开放CC BY-NC-SA
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[刘佳]'s Articles
Baidu academic
Similar articles in Baidu academic
[刘佳]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[刘佳]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.