| 基于动态流场数据的虚拟港口建模方法 |
| |
| 引用本文： | 陈立家,王凯,魏天明,郝国柱.基于动态流场数据的虚拟港口建模方法[J].交通运输工程学报,2022,22(2):287-297. |
| |
| 作者姓名： | 陈立家 王凯 魏天明 郝国柱 |
| |
| 作者单位： | 1.武汉理工大学 航运学院，湖北 武汉 4300632.武汉理工大学 内河航运技术湖北省重点实验室，湖北 武汉 430063 |
| |
| 基金项目： | 国家重点研发计划2018YFC0810400国家重点研发计划2018YFC1407400 |
| |
| 摘 要： | 为了实现港口数字化升级，提出了一种基于动态流场数据的虚拟港口建模方法；采用三维重建模型从无人机倾斜摄影影像数据中重建了港口几何特征，获取高精度三维模型；引入了基于二次误差度量的边折叠算法简化模型，以避免数据量过大致使渲染效率低的问题；分析了欧拉法数值计算过程中的高耗时环节，建立了神经网络模型学习流场演化特征，加速投影项计算得到实时变化的流场数据，通过流场数据驱动水流动态渲染，结合光滑粒子流体动力学方法表现水流与船舶、陆地的交互动态，在保证渲染实时性的同时，提高渲染真实感。研究结果表明:重建的港口三维重建模型顶点数量可达3 320 937个，重建的网格模型在Meshlab中渲染频率为78.7 Hz；经过模型简化降低90.0%的模型顶点数量后，模型顶点数量缩减为332 836个，渲染频率提升至108.7 Hz，模型简化后几何误差小于2.0%；在256×256的流场网格下，采用神经网络加速的网格流体计算方法所得水流速度场平均更新间隔约为17 ms，平均仿真精度为88.6%；通过开源图像引擎驱动流场数据和港口三维模型，平均渲染频率可达50.5 Hz。可见，该方法可有效解决高精度实时渲染中的关键问题，以达到仿真精度与渲染效率间的动态平衡，在精度损失较小的情况下实现较高精度的虚拟港口建模与实时动态仿真。
|
| 关 键 词： | 虚拟港口 三维重建 神经网络 倾斜摄影 水流模拟 |
| 收稿时间： | 2021-10-13 |
Virtual port modeling method based on dynamic fluid field data |
| |
| CHEN Li-jia,WANG Kai,WEI Tian-ming,HAO Guo-zhu.Virtual port modeling method based on dynamic fluid field data[J].Journal of Traffic and Transportation Engineering,2022,22(2):287-297. |
| |
| Authors: | CHEN Li-jia WANG Kai WEI Tian-ming HAO Guo-zhu |
| |
| Affiliation: | 1.School of Navigation, Wuhan University of Technology, Wuhan 430063, Hubei, China2.Hubei Key Laboratory of Inland Shipping Technology, Wuhan University of Technology, Wuhan 430063, Hubei, China |
| |
| Abstract: | To achieve the digital upgradation of ports, a virtual port modeling method based on the dynamic flow field data was proposed. The geometric feature of a port was reconstructed by the three-dimensional (3D) reconstruction model with the aerial image data of unmanned aerial vehicle (UAV) oblique photography, and a high-precision 3D model was established. The simplified model of the edge collapse algorithm based on the quadric error metric was introduced to prevent low rendering efficiency caused by the data overflow. The high time consuming step in the numerical calculation of Euler method was analyzed. A neural network model was built to learn the evolution feature of the flow field. The calculation of the projection term was accelerated to produce the dynamic flow field data that were used to drive the dynamic rendering of water flow. The smoothed particle hydrodynamics (SPH) method was employed to reflect the interactions of water flow with ships and land. In this way, not only was the real-time performance of rendering ensured, but also the realistic effect of rendering was improved. Research results show that the 3D reconstruction model of the reconstructed port has 3 320 937 vertices, and the rendering frequency of the reconstructed grid model is 78.7 Hz in Meshlab. After almost 90.0% of the vertices are removed from the model via model simplification, the number of vertices reduces to 332 836, and the rendering frequency enhances to 108.7 Hz. The geometric errors of the model are smaller than 2.0% after simplification. In a 256×256 flow field grid, the average update interval is roughly 17 ms for the water flow velocity field obtained by the grid fluid calculation method accelerated by a neural network, and the average simulation precision is 88.6%. When the flow field data and 3D port model were driven by an open scene graph (OSG) engine, the average rendering frequency can reach 50.5 Hz. In conclusion, the proposed method can effectively solve the key problems in high-precision real-time rendering to achieve the dynamic balance between simulation precision and rendering efficiency. It enables high-precision virtual port modeling and real-time dynamic simulation without great precision loss. 1 tab, 11 figs, 30 refs. |
| |
| Keywords: | |
|
| 点击此处可从《交通运输工程学报》浏览原始摘要信息 |
|
点击此处可从《交通运输工程学报》下载全文 |
|
