Matching and 3-D Reconstruction of Multibubbles Based on Virtual Stereo Vision

Abstract

3-D characteristics of bubbles are of great importance to understand the flow mechanism in bubbly flows. A complete matching method based on virtual stereo vision is presented in this paper, and 3-D motion parameters are reconstructed. According to the captured images of multiple bubbles from two virtual views, a series of image processing steps are designed to extract the feature parameters of bubbles for the following work. To establish the correspondence relationship of the identical bubble from the two half images, a stereo matching method based on a multivariable constrained threshold is proposed to figure out the identical bubble from two views. On this base, a 3-D polar coordinate homonymy correlation algorithm is developed to keep the correspondence of the multibubbles from successive frames. Combined with the calibration, 3-D multibubble trajectories and velocities are then reconstructed effectively. The reconstruction accuracy of the measurement distance between the bubbles is finally verified by capturing the image of the calibration reference. The results show that the stereo matching method and the tracking algorithm of bubbles are valid, and with high precision.