基于深度相机Kinect的植物叶片重建研究

发布时间：2018-01-06 17:22

本文关键词：基于深度相机Kinect的植物叶片重建研究　出处：《江苏大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着计算机硬件技术和图形处理技术的进一步发展,虚拟现实技术越来越广泛运用于人们的日常生活中,包括科幻影视、体感游戏、智能穿戴等。而虚拟植物建模研究作为计算机图形学领域的重要组成部分,也一直在不断发展。从早期分形与L系统等经典植物建模方法,至基于图像的植物建模方法,再到现今基于点云的植物建模方法,模型在运行速度与真实感方面都得到了极大的提高。本文以植株的叶片作为研究对象,提出基于有限细节多密度点云加密植物叶片建模算法。首先通过深度相机Kinect扫描获得叶片信息,其中包含深度数据和颜色数据,并针对扫描中出现的一系列问题提出相关的研究方法,最终得到植物叶片模型。具体研究的相关内容为:(1)通过Kinect获得植物单叶片点云模型,在初始的去噪、匹配等预处理操作以后,针对叶片点云提出一种基于RGB的点云递减简化方法,该方法结合传统的包围盒算法能使简化后的点云颜色过度更自然,并且有效分离粘合叶裂。(2)结合隐式函数曲面重建原理,将基于RGB信息递减简化后的叶片点云采用Poisson曲面重建算法,结合原始点云的曲面重建,分析其相似程度、重建细节效果以及重建时间空间性能等。(3)在基于人眼视觉识别的局限性上提出一种有限细节多密度点云重建算法,与传统的网格重建算法不同,主要通过以点代面不断细化点的密度来产生视觉误差上的模糊曲面,实验证明其重建效果和速度在一定程度上优于网格重建。(4)将研究的内容以模块化的方式,结合OpenGL图形库和Qt GUI编程,构建出交互式的叶片点云简化、曲面重建、多密度重建的原型系统,并根据用户需求设置相关参数,更直观、更丰富地展现实验结果。
[Abstract]:With the further development of computer hardware technology and graphics processing technology, virtual reality technology is more and more widely used in people's daily life, including science fiction film and television, body sense games. As an important part of computer graphics research on virtual plant modeling has been developing continuously from the early fractal and L system classical plant modeling methods. To the image-based plant modeling method, and now the point cloud-based plant modeling method, the model has been greatly improved in terms of speed and realism. In this paper, the plant leaves as the research object. A modeling algorithm based on finite details and multi-density point clouds is proposed. Firstly, the leaf information is obtained by depth camera Kinect scanning, which includes depth data and color data. Finally, the plant leaf model is obtained. The specific research content is: 1) the single leaf point cloud model of plant is obtained by Kinect. After the initial de-noising, matching and other pre-processing operations, a point cloud decrement simplification method based on RGB is proposed for the blade point cloud. This method combined with the traditional bounding box algorithm can make the color of the simplified point cloud more natural, and can effectively separate the splits of the splits.) combined with the principle of surface reconstruction of implicit functions. The blade point cloud which is simplified by decreasing RGB information is reconstructed by using Poisson surface reconstruction algorithm, and the similarity degree is analyzed by combining the surface reconstruction of the original point cloud. Based on the limitations of human visual recognition, this paper proposes a finite detail multi-density point cloud reconstruction algorithm, which is different from the traditional mesh reconstruction algorithm. The fuzzy surface on the visual error is produced mainly by using the point surface to refine the density of the point continuously. Experiments show that the reconstruction effect and speed is better than that of mesh reconstruction to a certain extent.) the contents of the study are modular, combined with OpenGL graphics library and QT GUI programming. An interactive prototype system of point cloud simplification, curved surface reconstruction and multi-density reconstruction is constructed, and the relevant parameters are set up according to the user's needs. The experimental results are more intuitive and abundant.
【学位授予单位】：江苏大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】