PIV/BOS synthetic image generation in variable density environments for error analysis and experiment design
PIV/BOS synthetic image generation in variable density environments for error analysis and experiment design
复制标题
DOI：
10.1088/1361-6501/ab1ca8
复制
发表时间：
2017-01
影响因子：
2.4
通讯作者：
Lalit K. Rajendran;S. Bane;P. Vlachos
中科院分区：
文献类型：
--
作者：
Lalit K. Rajendran;S. Bane;P. Vlachos
我们提出了一种基于光线跟踪的图像生成方法，可用于绘制逼真的粒子图像测速（PIV）和背景定向纹影（BOS）实验中存在的密度/折射率梯度的图像。这种方法能够模拟气动热力学实验的实验设计，误差和不确定性分析。通过从颗粒或点图案发出光线并将其传播通过密度梯度场和光学元件直到相机传感器来生成图像。渲染的图像是真实的，并且可以复制给定实验设置的特征，如光学像差和透视效果，这些可以故意引入用于误差分析。我们证明了这种方法，通过模拟一个已知的密度场的均匀浮力驱动的湍流直接数值模拟得到的BOS实验，并比较光线跟踪的光线位移从BOS理论的结果。光线位移与参考数据吻合良好。这种方法提供了一个框架，进一步开发的模拟工具，用于实验设计和开发的图像分析工具的PIV和BOS应用。在Python CUDA程序中实现所提出的方法可作为研究人员的开源软件。
We present an image generation methodology based on ray tracing that can be used to render realistic images of particle image velocimetry (PIV) and background oriented schlieren (BOS) experiments in the presence of density/refractive index gradients. This methodology enables the simulation of aero-thermodynamics experiments for experiment design, error, and uncertainty analysis. Images are generated by emanating light rays from the particles or dot pattern, and propagating them through the density gradient field and the optical elements, up to the camera sensor. The rendered images are realistic, and can replicate the features of a given experimental setup, like optical aberrations and perspective effects, which can be deliberately introduced for error analysis. We demonstrate this methodology by simulating a BOS experiment with a known density field obtained from direct numerical simulations of homogeneous buoyancy driven turbulence, and comparing the light ray displacements from ray tracing to results from BOS theory. The light ray displacements show good agreement with the reference data. This methodology provides a framework for further development of simulation tools for use in experiment design and development of image analysis tools for PIV and BOS applications. An implementation of the proposed methodology in a Python-CUDA program is made available as an open source software for researchers.