Using the watershed method in determining the size of intersecting particles of crushed grain

Abstract

The determination of the granulometric composition of ground grain is of great importance when feeding farm animals and poultry. The sieve method of determining the granulometric composition is traditionally used. With the development of modern digital technologies, it has become possible to determine the granulometric composition of grain without using special sieves. We have proposed a method for determining the particle sizes of crushed grain using color clustering and morphological segmentation. However, this method does not solve the problem of particle overlap, when several contacting components are perceived as one large particle, which ultimately affects the final result of the analysis and gives a serious error relative to the traditional sieving method. In this regard, it was decided to improve the method by additional image analysis using the watershed method. The article presents the results of estimating the particle sizes of crushed grain when they intersect and overlap each other. The sizing technique consisted in the implementation of the following operations: image acquisition; entropy filtering of a grayscale image; calculation of the average color value for each color cluster of the image; construction of a refined binary mask; calculation of the gradient representation of the image with grains; calculation of foreground markers; calculation of background markers; removal from the gradient representation of the image all local brightness minima outside the constructed markers; application of the watershed method; determination of the area and maximum diameter of the Feret and sorting of all areas by size. As a result of the analysis of images of crushed grain particles, histograms of the particle distribution by their areas and lengths were constructed according to the proposed method. The complexity of estimating the size of crushed grain by existing classical methods has been revealed and it has been established that the watershed method gives a sufficiently high error (about 32 %) in determining the particle area, but can be used to determine the linear particle sizes through which their equivalent diameter can be expressed.

Keywords

• granulometric composition
• diameter of the feret
• image processing
• intersection
• area
• background
• color segmentation
• color space