Automatic spatial characterization of low-speed streaks from thermal images

Abstract Flow visualization is an important tool for investigating turbulent flow, and, specifically, for characterizing low-speed streaks in the boundary layer. The span-wise spatial characteristics of these streaks are commonly extracted by human visual inspection, which is time consuming and subject to human errors and biases. Attempts to develop automatic methods have relied exclusively on spectral techniques, using mostly the autocorrelation or its Fourier transform, the spatial spectrum. However, the autocorrelation tends to get flattened with the amount of data analyzed and has been reported to provide biased estimates. Furthermore, it estimates only the mean spacing and does not provide a direct measure of its distribution. In this paper, an alternative automatic method is developed based on edge detection, and is applied to thermal images obtained by infrared thermography of a heated wall exposed to a turbulent flow. The method presented yields not only the spacing between the low-speed streaks but also their width and separation. The analysis indicates that the spacing (120 ± 52 wall units) is divided almost evenly between the width (65 ± 33 wall units) and the separation (55 ± 40 wall units) between the streaks, and that the width and separation are statistically independent. We also present a statistical model for the data, and demonstrate that when the spatial parameters of the streaks are so widely distributed, the spectral methods are not reliable.