Hough transform based method for air bubbles and oil drops segmentation in dispersions occurring in stirred bioreactors

Industrial production of important fermentation products such as enzymes, organic acids, antibiotics, and aroma compounds, amongst others, involves the dispersion of various phases. Therefore, it is important to determine the chemical and physiochemical conditions under which the culture achieves the highest yields and production. The measurement of the size distribution of air bubbles and oil drops in a fermentation process allows evaluating the interfacial area, and therefore, to infer the transfer efficiency of nutrients to the microbial culture. The automatic on-line analysis of multiphase dispersions occurring in microbial cultures in mechanically stirred bioreactors, presents a number of important problems for image acquisition and segmentation, including heterogeneous transparency of moving objects of interest and background, blurring, overlapping and artifacts. This work proposes a improved Hough transform-based method that allows to quantitatively evaluate the size distributions of air bubbles and oil drops in a mechanically stirred bioreactor, in a more efficient and 55% less time-consuming way.