A Collision Detection Method of Particle and Bubble Based on Otsu Image Segmentation

"In order to accurately detect whether the collision occurs between particle and bubble in flotation process, a collision detection method of particle and bubble, which based on morphological image processing and Otsu method, is presented. First of all, it utilizes image pre-processing to optimize images, the target area and the background area are segmented by Otsu method. Then, particle and bubble are fitted to a circle combined with the least-square method. It’s convenient to obtain the coordinate information and radius. Finally, the position of all particles is extracted to an image, which achieves the particle tracking and avoids the complexity of the continuous frame image tracking.Experimental analysis shows that the relationship between the two ordinate difference d and the bubble radius r has a certain influence on the collision detection when the dripping particles are close to the bubble, when d>r, the detection accuracy is 92%;When d=1/2r, the detection accuracy is 97%;When d=0, the detection accuracy is 100%. The result indicates that this algorithm can detect the collision between particle and bubble quickly, and the working hours of manual test are saved effectively ,the best position of dripping particles close to the bubble which is taken with a CCD camera should be in the range of the particle and bubble’s ordinate difference less than half of the bubble radius.INTRODUCTIONFlotation is an important technical means to separate useful minerals from ores. The flotation process of ores and materials in the world is treated by flotation technology up to billions of tons every year, so the study of flotation process is particularly important. Raising the collision probability between mineral particles and bubbles in flotation process is one of the key factors to improve flotation separation efficiency[1]. The study indicates that[2-4] the collision probability between particles and bubbles is influenced by many factors, such as particle density, particle size bubble size, etc. The analysis of these influential factors requires a lot of experiments. Therefore, it is of great significance to design an accurate collision detection method for the research of this field."