Imaging and Post-Processing of Early Combustion in a Spark- Ignition Engine via Endoscopic Access.

Two different high-speed cameras, a Photron SA-Z and a Vision Research Phantom v7.3, were used together with a large-aperture endoscopic system to visualize the early flame in a nearly unmodified production IC engine. At very low light levels, the patterned read-out noise on the Phantom v7.3 becomes significant. Filtering in the Fourier domain was effective in suppressing this noise component to acceptable levels. A previously-developed algorithm with automatic dynamic thresholding was expanded to separately detect spark ignition and flame kernel in the image sequences. The robustness of the algorithm was examined by processing data sets from different engine operating conditions as well as at different levels of lens collection efficiency (f-number) for the same engine operating point. [ABSTRACT FROM AUTHOR]