Quantitative comparison of image analysis methods for particle mixing in rotary drums

Optical imaging is a powerful tool for the study of particle mixing dynamics in rotary drums because it offers abundant choices of image analysis methods. However, little effort has been made to investigate the compatibility of the results obtained by those different methods, and the magnitude of deviations between them remains unclear. The present work contributes to quantitative comparison of two classic image analysis methods (Variance and Contact), with special attention to the procedures for measuring the mixing time which is an important parameter describing the mixing efficiency of drum equipments. For conditions tested in our experiment (red/white binary mixing in a drum of ID 206 mm captured by a camera with 25 fps), a relative deviation of 20%-36% was found between mixing times determined by different image processing algorithms and regression models. Compared to Variance, Contact can improve the measuring accuracy by 20% but requires 40% more computing time. An increase of the number of cells shows no significant effect on the processing speed of Variance. For regression analysis of the data, the second-order model is shown to be more reliable than the first-order model. Based on the comparative results of the present work, we would suggested that image analysis procedures for particle mixing be standardized so that reasonable conclusions can be drawn from measurements of different resources. (C) 2014 Elsevier B.V. All rights reserved.