基于反射率的多尺度水面重建方法  

Research on multi-scale reconstruction of water surfaces based on reflectance

作　　者：沈亮[1,2] 王昱哲 张娟[1,2] 朱登明 石敏[3] 

机构地区：[1]中国科学院计算技术研究所虚拟现实实验室,北京100190 [2]中国科学院大学,北京100049 [3]华北电力大学控制与计算机工程学院,北京102206

出　　处：《中国科学：信息科学》2017年第10期1369-1380,共12页Scientia Sinica(Informationis)

基　　金：国家自然科学基金(批准号:61173067;61379085;61532002;61300131);国家高技术研究发展计划(863计划)(批准号:2015A A016401)资助项目

摘　　要：为了生成逼真的水面动画,以单摄像头拍摄的水面视频作为输入数据,本文提出一种基于反射率的多尺度水面重建方法.在传统SFS方法的基础上构建了加入反射率条件的模型,求解整体水面高度场数据,并利用局部优化算法使得高度场对波浪的拟合更加准确,得到真实生动水面形态.针对单帧求解中波浪跳变的问题,利用浅水方程求解高度场中点的速度信息,根据各点的速度在时域上对求得的水面动画进行平滑.最后提出了一种高效实用的波浪翻滚重建算法,可根据水面实际情况,利用求得的速度信息和高度场的梯度数据再现波浪的翻滚效果,使得波浪更加生动.结果表明,使用该方法可以得到具有真实感的水面动画视频,满足各项工业领域用户的需求.Aimed at reconstructing vivid water animations,we present a multi-scale reconstruction framework based on reflectance using video from one camera as input data. Based on traditional the shape from shading method,we build our reflectance model to reconstruct large-scale wave trend information. We then use a local algorithm to reconstruct the details of waves. By combining the two results,we can construct a more accurate height field for the water surface. To solve the problem of unsmoothed animation frames,we use shallow water equations to calculate the velocity field of the water surface. We then optimize the flatness between frames using velocity data. Finally,we propose an efficient algorithm for the reconstruction of wave tumbling based on the gradient of the height field and speed of the water surface. Results demonstrate that our method can effectively synthesize large-scale water animations with a high sense of reality,which meets user demands in many industries.

关 键 词：水面 重建 数据驱动 浅水方程 反射率 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]