Template-based bubble identification and tracking in image sequences

Abstract

We have developed a template-based system for identifying and tracking vapour bubbles. The bubble images are acquired from a high-speed digital camera. Each image sequence is recorded digitally for off-line image processing. The bubble detection process is organized in two phases: bubble identification and bubble tracking. In the first phase, a principal component analysis is applied for detecting bubbles roughly. After the detection, some bubbles' sub-images are used as templates in order to identify the bubble contours accurately. The prerequisite of using this system is that the bubbles are assumed to be of elliptic shape. In the second phase, the templates are used again for tracking bubbles. The modified Newton–Raphson algorithm is applied to reduce the search time. The advantage of this method is that it can recognize bubbles that are recorded in a wide-range of experimental environments such as varied illumination, different bubble appearances and so on. After the bubble identification, some parameters such as bubble size, position, moving speed, and the bubble nucleation sites are obtained. They are important for the research on heat transfer.