Arbitrarily Three-dimensioanl Boundary Recognition Based Threedimensional PIV Measurement Technique For Complex Near-wall Flow Field

Fluid-structure interaction(FSI) phenomena are very widespread in the study of nature and engineering practice, such as a swimming fish can reduce the drag force by changing its body and blades in an engine would vibrate intensely with high-temperature air. In complex near-wall flow filed, calculating 3D-3C velocity field accurately is very significant for solving problems in these phenomena. Particle image velocimetry techniques are widely used in experimental fluid mechanics because of many advantages. However, traditional PIV cannot distinguish the solid boundary and tracer particle in PIV images of near-wall flow field and this can increase the 2D cross-correlation calculation error. Moreover, the interface between fluid and solid is an arbitrarily 3D moving boundary in FSI experiments, however, 3D boundary recognition algorithms are still lacking in PIV measurement. Therefore, in the present work, an algorithm of three-dimensional solid boundary recognition is proposed to accurately determine the 3D location of the interface between fluid and solid in nearwall flow field measurement. A smart algorithm based on SURF(Speeded up robust features) algorithm is employed to determine the solid boundary information in near-wall flow measurement. In addition, a state-of-the-art Tomographic PIV algorithm based on MLOS-SMART three-dimensional particle field reconstruction and 3D cross-correlation algorithms is developed to calculate the 3D-3C velocity field in near-wall flow filed. Eventually, the 3D solid interface feature and near-wall 3D-3C velocity field are both obtained by newly developed algorithms.Firstly, a state-of-the-art 3D boundary recognition algorithm based on SURF algorithm is proposed:(1) detect keypoints in the images of two different viewpoints, determine the feature descriptors for every keypoint and match these feature descriptors by nearest-neighbor matching algorithm and epipolar constraint;(2) triangulate these matches based on two camera calibration matrices and reconstruct the 3D point cloud of these matches, moreover, reconstruct the 3D surface by Delaunay triangulation algorithm. Furthermore, the 3D boundary recognition algorithm is verified by using a set of circular cylinder images where the exact curvatures are known.Moreover, the 3D velocity field of a near-wall circular cylinder wake is calculated by CFD and used for generating synthetic particle image sets by applying four camera calibration matrices. 3D boundary geometric information of the circular cylinder and 3D-3C velocity field around the solid boundary are obtained by applying newly developed image processing algorithms to these synthetic particle image sets. Finally, the validity and accuracy of the algorithms are verified by processing the synthetic images and comparing the calculated velocity filed and boundary geometry with CFD data and the exact cylinder dimensions. These developed algorithms will provide a new experimental measurement technique for complex near-wall flow field and FSI problems.