Machine Vision Based Measuring Method For Particle Size In A Pan Coater

The pan coater is widely used in the pharmaceutical industries and food industries.It is an important process equipment for film coating of spherical particles such as pills and candies.In the coating process,the particle size is an important process information that needs to be continuously monitored.The traditional off-line manual sampling and measuring method has the disadvantages of a few samples,delayed measuring result,low efficiency,and possible contamination of the products(medicines or foods).This paper is focused on particle size measuring method based on machine vision,with the aim to improve the measuring accuracy and efficiency for particle size measurement in pan coaters,which is of theoretical and practical importance.Firstly,the BY600 pan coater filled with particles of different sizes was used as experimental platform.According to the structural characteristics of the pan coater and the motion behavior of the material,the machine vision system was designed,with proper selection of the light source,industrial camera,lens and a special supporting framework were selected.The designed hardware system was tested and the particle image acquisition was completed.Then,in order to deal with the image noise and the unclear particle edge,the raw image was processed by filtering,the morphological opening and closing reconstruction.In order to separate overlapping particles in the image,an improved watershed segmentation algorithm combining distance transformation and local maximum value image was adopted to reduce over-segmentation which usually occurs in image segmentation using traditional watershed algorithm.By combination of the watershed ridge diagram and the binary image of the particle,the shape of the particle can be maintained and each individual particle be effectively separated.This lays a good foundation for particle size measurement and particle size distribution(PSD)analysis.Finally,for the particle size measurement,the minimum circumscribed circle fitting method is used to measure the pixel diameter of individual particles and PSD of all particles in the image.In order to solve the problem that segmentation of overlapping particles greatly affects the measuring accuracy of particle size,adaptive fuzzy C-means clustering algorithm was adopted to divide the measured pixel diameters into serveral clusters.i With known possible particle size types,the corresponding particle size can be obtained.The experimental results show that the average measuring accuracy of over the proposed method is about 92.5% after adopting the clustering algorithm,and the measured particle size distribution map is consistent with the actual particle size distribution,which demonstrates the effectivenss of the proposed methodThe machine vision inspecting platform designed in the present work is simple to install,and the proposed algorithm for measuring the particle size in the pan coater has satisfactory accuracy.Results of this work provides a possible solution for automatic realtime detection of paticle size in pan coaters.