| 摘要: |
| 水体辐射传输方程是复杂的微积分方程,只能利用数值方法求解,如Monte Carlo光线追踪法、不变嵌入法、离散坐标法等,其中,Monte Carlo方法是目前解决水体水下光场三维问题的唯一有效方法.根据辐射传输理论,开发了水下光场的Monte Carlo模拟模型,主要包含大气、水-气界面、层化水体和水底边界4个模块.实现了模拟任意太阳角度、不同水体固有光学属性和任意深度条件下,考虑大气、粗糙水面和水底边界的水下光场,能够获取辐亮度、辐照度等辐射量的空间分布.该模型暂不考虑Raman散射、偏振、内部光源的影响.实现了GPU加速水下光场Monte Carlo模拟,并用Mobley等提出的海洋光学标准问题中的问题1~6进行验证.在两种计算环境下,通过对不同边界条件下的CPU、GPU运行时间及加速比的对比,发现GPU计算可以达到几百至上千倍的加速比. |
| 关键词: GPU 水体辐射传输方程 Monte Carlo模拟 水下光场 |
| DOI：10.18307/2016.0322 |
| 分类号: |
| 基金项目:国家自然科学基金项目(41471284,41431176)、国家高技术研究发展计划"863"项目(2012AA12A303)和国家重点基础研究发展计划"973"项目(2013CB733403)联合资助. |
|
| Accelerating Monte Carlo radiative transfer simulation of water using GPU technique |
|
DU Keping1, XUE Kun2,3
|
|
1.State Key Laboratory of Remote Sensing Science, School of Geography, Beijing Normal University, Beijing 100875, P. R. China;2.State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing 210008, P. R. China;3.University of Chinese Academy of Sciences, Beijing 100049, P. R. China
|
| Abstract: |
| Radiative Transfer Equations(RTE) of water body are complex integro-differential equations, which can be solved by different numerical methods, e.g., Monte Carlo ray tracing, invariant imbedding, and discrete ordinates. Monte Carlo method is a powerful technique, which can be used in any water body, even those whose boundary conditions and inherent optical properties(IOPs) vary in three dimensions. However, the Monte Carlo method is computationally costly, which limits the use for many problems in optical oceanography. In this paper, a new kind of acceleration technology to accelerate the ocean radiative transfer simulation, using the CUDA-enabled graphics processing unit(GPU) is presented. With the approach's help, it is easy to code on NVIDIA GPUs and there is no need to worry about the hardware details of a specific GPU. Firstly, some basic ideas of ocean radiative Monte Carlo simulation are introduced, then GPU programs for ocean radiative transfer simulation are implemented. Finally, the performances of the two GPUs(NVIDIA GTX 670 GPU and NVIDIA Quadro 6000 GPU) with their CPU counterparts are compared. From our numerical results, the speedup over hundreds of times for solving the issues is achieved compared with that obtained using CPU. |
| Key words: Graphics processing unit(GPU) radiative transfer equation Monte Carlo simulation underwater light field |
