A computer vision approach for analysis of detonation cellular structures

In the current study, we present a novel computer-vision-based method for automated detection, measurement, and statistical analysis of detonation cellular structure images. The new approach consists of four primary steps: (1) image preprocessing, (2) cell contour detection, (3) parameter optimization, and (4) statistical analysis. First, the cell size measurements from the proposed approach are extensively validated against other measurement methods for numerical soot foils. We demonstrate that the computer vision approach can measure the average cell dimensions with a maximum relative error of 30% for images with a very wide range of cell regularity levels and resolutions. For high-resolution regular and irregular patterned numerical soot foil images, the maximum relative errors decrease to 8% and 17%, respectively. Moreover, cell distribution histogram analysis is carried out for cases with irregular cellular structures. We show that the suggested method can capture the correct cell size distributions with reasonable accuracy in comparison with other measurement methods. Finally, we demonstrate the new computer vision approach capability to automatically analyze high-quality experimentally-derived detonation cellular structure images.