基于Android平台作物3D模型渲染方法的实现

发布时间：2018-03-20 19:19

本文选题：Android　切入点：三维点云重建　出处：《西北农林科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：作物三维模型渲染是农业信息化领域的研究热点。目前,作物的三维模型渲染多数是基于PC端。但在智能设备迅速发展的今天,人们希望能够在移动端有良好的视觉体验,因此本文通过对一些渲染算法移植,从而实现这一目标。其中的关键问题是如何将这些算法进行改进使其适应移动端低带宽、低功耗的需求。本研究以作物三维模型为研究对象,在详细分析其原理、实现过程后,基于Android平台结合OpenGL ES图形库实现了作物三维模型的渲染。主要内容如下:(1)实现了作物三维点云的重建算法,获取了三维模型。针对本设计输入的点云是精简去噪的点云,且其来自一个扫描设备或多个扫描设备的特点,选取适合该三维点云的重建算法—贪婪投影算法对玉兰树和玉米植株等多种作物进行重建,获得了很好的重建效果。通过与泊松重建算法对比,针对植物等散乱叶片的点云数据,虽然贪婪投影算法效率低于泊松重建算法约20.61%,但是能够体现植物的拓扑结构,而泊松重建虽然时间效率高,但出现冗余面。(2)实现了移动端三维模型的显示。针对三维模型如何在移动端显示的问题,本文采用了应用广泛的3D标准文件类型—STL文件保存获取到的三维模型。使用OpenGL ES在Android Studio上实现三维模型的获取,实验结果表明该读取方式可以对STL文件读取并显示,且显示位置自适应于手机屏幕。(3)实现了以光照为核心的移动端三维模型的渲染。针对模型不够真实,视觉效果不佳的问题,本文对模型进行了局部光照处理和全局光照处理。局部光照处理采用Phong氏光照模型,很好的模拟了高光效果;全局光照处理采用光线跟踪算法,很好的模拟了光照下的阴影效果。在网格面达到85436时,移动端依然可以实现渲染,渲染时间为36.24分。本文还对该方法进行了功能性测试和普适性测试。实验基于斯坦福Bunny密集点云数据对该方法进行功能性测试,选取八组数量不同的点云数据进行渲染对比,测试用例表明模型表面的光滑程度会对渲染结果造成影响,当点云数量少于3305时,高光面有明显的缺损;选取不同类型的三维模型进行普适性测试,测试用例表明该方法可以实现多种模型的渲染。通过以上两方面的测试,结果表明该方法实现了预期的功能,且能够广泛使用。
[Abstract]:Crop 3D model rendering is a hot topic in the field of agricultural informatization. At present, most of crop 3D model rendering is based on PC. However, with the rapid development of intelligent devices, people hope to have a good visual experience on mobile side. The key problem in this paper is how to adapt these algorithms to the low bandwidth on the mobile side. In this study, the three dimensional model of crop is taken as the research object, after the detailed analysis of its principle, the realization process, Based on Android platform and OpenGL es graphics library, the rendering of crop 3D model is realized. The main contents are as follows: 1) the reconstruction algorithm of crop 3D point cloud is realized, and the 3D model is obtained. The point cloud input in this design is a reduced denoising point cloud. Based on the characteristics of one or more scanning devices, the greedy projection algorithm, which is suitable for the 3D point cloud reconstruction, is selected to reconstruct magnolia and maize plants. Compared with Poisson's reconstruction algorithm, the greedy projection algorithm is less efficient than Poisson's reconstruction algorithm, but it can reflect the topological structure of plants, although the greedy projection algorithm is less efficient than Poisson's reconstruction algorithm. Poisson reconstruction has high time efficiency, but redundant surface. 2) realize the display of 3D model of mobile terminal. In view of the problem of how to display 3D model on mobile side, This paper adopts 3D standard file type -STL file which is widely used to save the acquired 3D model. We use OpenGL es to obtain 3D model on Android Studio. The experimental results show that the method can read and display the STL file. And the display position adapts to the mobile phone screen. 3) it realizes the rendering of the 3D model of the mobile end with illumination as the core. Aiming at the problem that the model is not real enough and the visual effect is not good, In this paper, the local illumination processing and the global illumination processing are carried out. The Phong illumination model is used in the local illumination processing, and the highlight effect is well simulated, and the global illumination processing is based on the ray-tracking algorithm. Very good simulation of the shadow effect under light. When the mesh surface reaches 85436, the mobile side can still render, The rendering time is 36.24 points. This paper also carries on the functional test and the universality test to this method. The experiment carries on the function test based on the Stanford Bunny dense point cloud data, selects eight groups of points cloud data to render the contrast, Test cases show that the smoothness of the surface of the model will affect the rendering results. When the number of point clouds is less than 3305, the high light surface has obvious defects. The test cases show that the method can be used to render multiple models. The results show that the proposed method achieves the expected function and can be widely used.
【学位授予单位】：西北农林科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP316

【参考文献】