At present, the particle size of compound in sintering production is mainly obtained by manual inspection, this operation mode is not continuous and accuracy needs to be improved, and it is difficult to determine the quantitative relationship between the particle size distribution and the sintering production parameters. Therefore, a particle size detection model of sintering compound is proposed, which adopts CCD industrial camera and industrial light source as the main acquisition equipment of sintering compound image; in the image pre-processing process, the weighted average method is used for the graying of the image; in the construction of the particle size detection model, the Unet network is applied to the segmentation of the sintering compound image. The results show that the use of industrial light source can improve the brightness when collecting, and also reduce the influence of external light to the greatest extent out, so as to ensure the stability of sintering compound image. Image preprocessing is helpful to distinguish the particle characteristics of sintering compound. On this basis, median filtering and histogram equalization are more suitable to characterize the edges of sintering compound particles in the image and blur the internal characteristics of particles, which have a good effect on removing noise and other unfavorable factors. The trained Unet network segmentation model for sintering compound has segmentation accuracy of more than 91% and segmentation accuracy error of less than 9%, which has a good segmentation effect on the compound image. The application of this model can provide timely and accurate particle size distribution data for sintering production, help improve the efficiency of particle size detection of sintering compound, and help enterprises improve economic and social benefits.