Automated segmentation and characterization of ion-abrasion scanning electron microscopy fuel cell images

This paper presents a method for semi-automatically segmenting ion-abrasion scanning electron microscopy (I-A SEM) images of fuel cell anodes into discrete regions representing the metal, ceramic, and empty space components of a given volume of a fuel cell anode. Segmentation is accomplished by the use of the fast marching method (FMM) proceeding from multiple automatically-generated seeds. Statistical region merging (SRM) is used to consolidate similar segments. Final consolidation of regions is accomplished using simple global thresholding of the merged regions and manual correction, if necessary. Following segmentation of a stack of images representing a complete anode, we convert the fuel cell volume into a connected graph representation. Characterization of many of the fuel cell's geometric properties can then be accomplished by graph algorithmic methods. As an example, we find the proportion of active and inactive triple-phase boundary (TPB) points using an A* graph search algorithm.