To conduct the core image particle size analysis,a semi-automatic segmentation method based on K-means clustering algorithms was proposed and a set of grain size image processing programs was formed.In the image processing,a superpixels algorithm was applied to process the polarizing microscopic image of cores,and the excessive segmentation results were achieved.Then the K-means clustering and regional integration were conducted on the segmentation results.In this way,the speed of operation was greatly improved,and the edge information of images was utilized to obtain more reasonable results.Based on the VB.NET 2008 platform,a semi-automatic software,being prove with the functions of image acquisition,image processing,analysis of grain size parameters,types of gravel classification and measurement reporting,was built according to the proposed algorithm.The method greatly improves the efficiency of the core particle size analysis.