山脉图像轮廓线引导的三维虚拟地形合成

【作者】 李伟；

【导师】 庞明勇；

【作者基本信息】 南京师范大学 ， 教育技术学， 2018， 硕士

【摘要】 随着智能信息技术的发展,虚拟现实技术将为探索未来教育教学新模式提供支持。为了生成与真实室外环境高度近似的数字化环境,获得亲临对应真实环境的体验,则离不开真实感强的三维虚拟地形。因此,三维虚拟地形合成已成为虚拟现实等领域研究的一个热点。本文围绕三维虚拟地形合成技术这一课题进行研究,提出了由山脉轮廓线引导的三维虚拟地形合成算法,主要内容有山体轮廓线提取、全局特征草图构建、初始地形合成、高程匹配等方面。本文主要工作如下:1.本文算法以一幅真实山脉图像为输入,首先利用边缘检测技术对图像进行轮廓线检测,通过轮廓线拆分操作获得相互独立的山体轮廓线。2.在获取了独立的山体轮廓线之后,本文借助卷积神经网络模型预测每一条山体轮廓线的深度信息。然后根据预测出的深度信息,对山体轮廓线进行如下优化处理:轮廓线深度拟合、轮廓线深度排序、轮廓线间距调整以及轮廓线补全。最后,本文结合拟合后的轮廓线的深度信息、轮廓线间的遮挡关系和图像成像原理,对轮廓线进行三维重构,重构的轮廓线为控制三维虚拟地形合成的全局特征草图。3.为了使合成的三维虚拟地形更具真实感,本文方法在全局特征草图的控制下,以具有局部细节特征的地形块为基本元素,通过地形块匹配、填充及融合等操作合成初始的三维虚拟地形。4.由于地形块与轮廓线在局部细节特征上存在差异,使得合成的初始地形与山体轮廓线在高程上存在不匹配的情况,所以本文以轮廓线为约束条件,对合成的初始三维虚拟地形通过高程匹配处理,使最终生成三维虚拟地形与轮廓线相匹配。在上述算法理论研究基础上,本文实现了由山脉图像轮廓线引导的三维虚拟地形合成。实验结果表明:算法能够在真实山脉轮廓线的约束下合成结果可预期的真实感三维虚拟地形。更多还原

【Abstract】 With the development of intelligent information technology,virtual reality technology will become a new model of education and teaching in the future.In order to generate a digital environment that is highly similar to the real outdoor environment for obtaining the strongly realistic experience,it is inseparable from the photorealistic rendering of 3D virtual terrain.Therefore,3D terrain synthesis has become a hot spot in the field of virtual reality.In this paper,we focus on the topic of 3D terrain synthesis technology and propose a 3D terrain synthesis algorithm guided by outlines of mountain image.Our main works include mountain outlines extraction,global feature sketch construction,initial terrain synthesis and elevation matching as follows:1.The edge detection technology is used to detect the outlines of the image,and the independent outline of mountain is obtained through the outline splitting operation.2.After obtaining independent outlines of the mountain,we predict the depth information of each mountain outline by the deep of convolution neural network model.Then according to the predicted depth information,the following optimization process is executed on the outline:outline depth fitting,outline depth ranking,outline spacing adjustment and outline complementation.Finally,we reconstruct the outline by using the depth information of the fitted outline,the occlusion relationship between the outlines and the image imaging principle.The reconstructed outline is a global feature sketch that controls the synthesis of the 3D terrain.3.In order to make the synthesized 3D terrain more realistic,under the control of the global feature sketch,our algorithm takes the terrain block with the local detail as the basic element,synthesizes the initial 3D terrain by the terrain block matching,filling and blending.4.Due to the differences in the local detail between the terrain blocks and the outlines,the synthesized initial terrain and the mountain outline do not match on the elevation.Therefore,we take the outline as the constraint condition to analyze the synthesized initial 3D terrain through the elevation matching process;the final generated 3D terrain is matched with the outline.On the basis of the theoretical study of the algorithm above,we realize the 3D terrain synthesis guided by the outline of the mountain image.The experimental results show that our algorithm can synthesize desired 3D terrain under the constraints of real outline of mountains.更多还原