Particle size distribution is extremely important in the coal preparation industry. It is traditionally analysed by a manual screening method, which is relatively time-consuming and cannot immediately guide production. In this paper, an image segmentation method for images of coal particles is proposed. It employs the watershed algorithm, k-nearest neighbour algorithm, and convex shell method to achieve preliminary segmentation, merge small pieces with large pieces, and split adhered particles, respectively. Comparing the automated segmentation using this method with manual segmentation, it is found that the results are comparable. The size distributions obtained by the automated and manual segmentation methods are nearly identical, and the standard deviation is less than 3%, indicating good reliability. This automated image segmentation method provides a new approach for rapidly analysing the size distribution of coal particles with size fractions defined according to consumer requirements.

粒度分布在选煤行业中极为重要。传统上，它是通过手动筛选方法进行分析的，该方法比较耗时且无法立即指导生产。本文提出了一种煤颗粒图像的图像分割方法。它采用分水岭算法k近邻算法和凸壳方法来实现初步分割，将小片段合并为大片段，并分别拆分附着的粒子。将使用此方法的自动分割与手动分割进行比较，发现结果是可比较的。通过自动和手动分割方法获得的尺寸分布几乎相同，标准偏差小于3％，表明可靠性高。这种自动图像分割方法提供了一种新方法，可以快速分析煤颗粒的尺寸分布，并根据用户需求定义尺寸分数。