U-Net-based image detection of the multi-particle mixing degree in a rotary drum

Mixing degree is a quantitative description of the uniformity of particle materials. This paper proposed an image detection method based on U-Net to calculate the mixing degree of multi-particles with four different sizes. First, mixing images of multi-particles are obtained from physical experiments; then the U-net is used to segment the image to extract the edge and area information of the particles; furthermore, an ellipse fitting method is used to get the particle center coordinates and particle radius information by using the edge and area information; finally, based on the above information, the Relative Standard Deviation can be calculated to describe the mixing degree of multi-particle. The test results show that the proposed method can detect the mixing degree of multi-particles in a rotary drum. And the mixing degree is varying with particle size due to the segregation phenomenon, which is consistent with the result in the literature. This method is promising in detecting the mixing degree of particle materials in the chemical industry.