Author(s): Daniel B. Bung, Daniel Valero
Linked Author(s): Daniel B. Bung
Keywords: Imaging techniques, computer vision, optical flow, turbulence, particle images
Abstract: Despite the continuous advances in numerical modeling methods and the continuous increase of computer power and storage capacities in the last decades, physical modeling remains a common technique for evaluation of flow processes. One major advantage of this classical approach is given by the transparency of experiments as the flow can be easily observed in the laboratory and effects of changing boundary conditions may be better understood. Nowadays, the flow is commonly captured in most experiments with means of digital cameras or high-speed cameras to get a better insight and for further analyses after the model test. Besides some well-known techniques, such Particle Image Velocimetry, some new methods can be found in Computer Vision disciplines allowing for more detailed investigation of the images. This paper presents two modules of the new open-source toolbox FlowCV which has been developed to be applied in hydraulic laboratories. The first module is given by a synthetic particle image generator which provides the user with particle images of a predefined motion and turbulence. The second module implements different Optical Flow methods for determination of obstacle movement (here: particle movement). The Farneb�ck method, which is presented in this paper, gives dense velocity fields, i. e. velocity data for every image pixel with moving particles. The Optical Flow results based on the synthetic images are then benchmarked against their predefined particle velocity. It will be shown that even the smallest turbulent flow structures are adequately detected
Year: 2017