关键词: 铁谱技术, 脉冲耦合神经网络, 数学形态学, 典型磨粒, 图像分割

Abstract: To improve the automation of ferrography, mathematical morphology is combined with Pulse Coupled Neural Network（PCNN）, then an automatic extraction method of wear particles for ferrographic images is proposed. This method employs synthesis color distance function to transform three-channel issue of color ferrographic images into single-channel issue. Thus, image segmentation is simplified. Then ferrographic images are indirectly segmented by simplified PCNN, and achieved binary images are processed utilizing mathematical morphology. With connected domain extraction algorithm of mathematical morphology, typical wear particles are extracted automatically. Experiments results show that compared with other methods, the proposed method can segment color ferrographic images effectively and realize automatic extraction of typical wear particles.

Key words: ferrography, pulse coupled neural network, mathematical morphology, typical wear particle, image segmentation