Distortion correction of two-component - two-dimensional PIV using a large imaging sensor with application to measurements of a turbulent boundary layer flow at $Re_{\tau} = 2386$

TL;DR

This study evaluates two distortion correction models for large-sensor PIV measurements of turbulent boundary layers, demonstrating that a bicubic polynomial model provides superior correction accuracy, with both correction approaches yielding comparable results.

Contribution

It introduces and compares two lens distortion correction models for high-resolution PIV measurements, highlighting the effectiveness of the bicubic polynomial model.

Findings

P3 dewarping model outperforms R2 in correction accuracy

Both correction approaches with P3 yield statistically similar results

Large sensors require distortion correction for accurate flow measurements

Abstract

In the past decade, advances in electronics technology have made larger imaging sensors available to the experimental fluid mechanics community. These advancements have enabled the measurement of 2-component 2-dimensional (2C-2D) velocity fields using particle image velocimetry (PIV) with much higher spatial resolution than previously possible. However, due to the large size of the sensor, the lens distortion needs to be taken into account as it will now have a more significant effect on the measurement quality that must be corrected to ensure accurate high-fidelity 2C-2D velocity field measurements. In this paper, two dewarping models, a second-order rational function (R2) and a bicubic polynomial (P3) are investigated with regards to 2C-2D PIV measurements of a turbulent boundary layer (TBL) using a large imaging sensor. Two approaches are considered and compared: (i) dewarping the images prior to the PIV cross-correlation analysis and (ii) undertaking the PIV cross-correlation analysis using the original recorded distorted images then followed by using the mapping functions derived for image dewarping to provide the correct spatial location of the velocity measurement point. The results demonstrate that the use of P3 dewarping model to correct lens distortion yields better results than the R2 dewarping model. Furthermore, both approaches for the P3 dewarping model yield results which are statistically indistinguishable.