Abstract:Ferrographic image segmentation is the premise of realizing ferrographic image automatic analysis.But ferrographic image contains both dark and bright wear particles,the background contains only a flat color and is different from the foreground colors,and adhering condition is existed among parts of the wear particles,so the effective segmentation of the ferrographic image is very difficult.According to the characteristics of ferrographic image,a method of ferrographic image segmentation was proposed,in which twice Kmeans clustering was used to extract particle area from ferrographic image,then improved watershed algorithm was used to segment the adhering wear particles from foreground area.The method solves the problem of wear particles’ incomplete extraction on the condition that light and dark particles are coexisting,and achieves the effective segmentation of adhering wear particles.Experimental results validate the feasibility of this method.