面向树木可视化的沉浸式虚拟场景的研究与实现

发布时间：2018-03-23 13:31

本文选题：树木可视化　切入点：虚拟现实技术　出处：《西北农林科技大学》2017年硕士论文

【摘要】：虚拟树木三维形态构建研究、树木可视化和沉浸式虚拟场景的构建是当前虚拟现实技术与农业信息化领域的研究热点和重点。穿戴式VR体验中稳定的图像质量、场景真实感、身体感应交互、环境变化连续性和社交等是影响沉浸感的主要因素。目前传统虚拟现实展示所带来的沉浸感有限,相关内容体验不够真实、模型不够仿真精细,树木可视化的效果不佳。基于规则与基于图像的方法存在操作复杂度较高,模型未能较好的达到仿真要求等问题,不能满足本研究的需求。为了改善以上问题,特使用真实树木点云三维数据,提高树木可视化的精细度。本文从模型重建、虚拟场景构建以及可穿戴式设备方面研究并实现一种实现虚拟现实系统的方法,主要的研究内容如下:(1)重建三维树木。针对树木模型构建不精确,不自然的问题,树木可视化工作基于三维激光点云结构模型的方法对树木进行重建。主要包括以下步骤:首先,利用点云数据,通过空间殖民算法生成树木骨架,对骨架使用广义圆柱体和管道模型原理绘制树木。其次,树木模型根据OBJ格式语法生成模型文件,使用Unity读取文件获得树木模型。树木材质通过PBR渲染模式与着色器生成,场景阴影的渲染使用逐像素处理的全局光实时照明模式。最后,本文的树木可视化具有高度的真实感、沉浸感效果。(2)基于Oculus与Unity的树木可视化沉浸式虚拟场景。由于显示设备和观察方式不同,VR漫游与相关设备关联使用Oculus VR设备,VR摄像机通过Oculus SDK开发。其主要内容包括摄像机运动、关联跟踪器、变换和渲染,可为使用者提供六自由度的观察体验。场景碰撞检测以改进的球形包围盒算法为基础,快速排除不可能碰撞的物体,再结合节省判断资源的方法,对树木主干的有效碰撞区域设置碰撞器,提升碰撞检测的效率。虚拟场景的搭建作为提升沉浸感的辅助措施,其中涉及到:天空、地面山脉、水面和静态植被。天空盒绑定摄像机,地面与山脉使用Terrain Engine和贴图复合。水体模拟使用投影网格和Gerstner波,静态植被通过3Ds max工具和Billboard原理构造。(3)系统优化与测试。本文通过对虚拟现实系统采用:DS管线渲染降低光照计算开销,剔除技术节省渲染资源以及LOD技术简化远距离模型的复杂度等方法优化多模型情况下系统的计算能力,从而保持系统流畅运行。本文最后使用功能用例的黑盒测试、志愿者系统体验等调查方式,统计反馈数据。与传统的VR展示方式对比,本虚拟现实系统测试结果表明:可达60FPS的较优流畅度,优秀率为70%,非常优秀率近30%;模型场景真实感反馈优秀率约为52.5%,非常优秀率近17.5%;树木可视化沉浸感体验优秀率为45%,非常优秀率近45%。通过以上结论判定,本文研究并实现的沉浸式系统可以满足课题预期的功能和性能要求。
[Abstract]:The research of virtual tree 3D shape construction, tree visualization and immersive virtual scene construction is the focus of research in the field of virtual reality technology and agricultural informatization. The image quality and scene reality in wearable VR experience are stable. Physical interaction, environmental continuity and social interaction are the main factors that affect immersion. At present, the traditional virtual reality display brings limited immersion, the related content experience is not real enough, the model is not accurate enough. The performance of tree visualization is not good. The rules based and image-based methods have some problems such as high operation complexity, the model can not meet the requirements of simulation, and can not meet the needs of this study. In this paper, we study and implement a method to realize virtual reality system from the aspects of model reconstruction, virtual scene construction and wearable devices. The main contents of the study are as follows: (1) Reconstruction of three-dimensional trees. Aiming at the imprecise and unnatural construction of tree models, Tree visualization is based on the 3D laser point cloud structure model, which includes the following steps: firstly, the tree skeleton is generated by using point cloud data and spatial colonization algorithm. Secondly, the tree model generates the model file according to the syntax of OBJ format, and uses Unity to read the file to obtain the tree model. The tree material is generated by PBR rendering mode and shader. The rendering of the shadow of the scene uses the global light real-time illumination mode which is processed by pixel by pixel. Finally, the tree visualization in this paper has a high degree of realism. Immersive effect. 2) based on Oculus and Unity, the visual immersive virtual scene of trees. The main content of the virtual scene is developed by Oculus SDK because of the different display devices and different viewing ways. The Oculus VR camera is developed by Oculus SDK. Including camera motion, Correlation tracker, transform and render, can provide the user with six degrees of freedom observation experience. Scene collision detection is based on the improved spherical bounding box algorithm, which can quickly eliminate objects that can not collide, and combine with the method of saving judgment resources. To improve the efficiency of collision detection by setting up collider in the effective impact area of tree trunk. The virtual scene is used as an auxiliary measure to enhance immersion, which involves: sky, mountain range on the ground, Water surface and static vegetation. Sky box binds camera. Ground and mountain use Terrain Engine and map composite. Water simulation uses projected grids and Gerstner waves. Static vegetation is optimized and tested by using 3Ds max tool and Billboard principle. Culling technology saves rendering resources and LOD technology simplifies the complexity of remote models to optimize the computing power of the system in the case of multiple models, so as to keep the system running smoothly. Finally, this paper uses the black box test of functional use cases. Compared with the traditional VR display method, the virtual reality system test results show that the optimal fluency of 60FPS can be achieved. The rate of excellence is 70, the rate of very good is nearly 30; the rate of model scene truthfulness feedback is about 52.5, and the rate of very excellent is nearly 17.5. the excellent rate of visual immersive experience of trees is 45, and the rate of very excellent is nearly 45.According to the above conclusions, The immersion system studied and implemented in this paper can meet the expected functional and performance requirements.
【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】：S758;TP391.9

【参考文献】