The paper proposes a method for the detection\nof bubble-like transparent objects in a liquid. The detection\nproblem is non-trivial since bubble appearance varies\nconsiderably due to different lighting conditions causing contrast\nreversal and multiple interreflections.We formulate the\nproblem as the detection of concentric circular arrangements\n(CCA). The CCAs are recovered in a hypothesize-optimizeverify\nframework. The hypothesis generation is based on\nsampling from the partially linked components of the nonmaximum\nsuppressed responses of oriented ridge filters,\nand is followed by the CCA parameter estimation. Parameter\noptimization is carried out by minimizing a novel\ncost-function. The performance was tested on gas dispersion\nimages of pulp suspension and oil dispersion images.\nThe mean error of gas/oil volume estimation was used as a\nperformance criterion due to the fact that the main goal of\nthe applications driving the research was the bubble volume\nestimation. The method achieved 28 and 13 % of gas and\noil volume estimation errors correspondingly outperforming\nthe OpenCV Circular Hough Transform in both cases and the\nWaldBoost detector in gas volume estimation.
Loading....