A hybrid method to segment the pores and throats of micromodels

Abstract

Micromodels made of silicon wafer are widely used in microfluidic system or oil displacement visual experiment. In this study, we develop a technique to partition the pores and throats of 2D porous micromodels, based on which the porous media structure could be described quantitatively. First, two series of medial axis are obtained based on the watershed algorithm and thinning algorithm, respectively. The thinning medial axis can be divided into four parts, i.e., node, endpoint, trunk, and branch, while the watershed medial axis does not have an endpoint. Then, the segmentation of pores and throats is initially carried out based on the watershed medial axis, and the interface of pore and throat is identified using the wavelet denoising method. Subsequently, the dead ends are further segmented, induced by the endpoints of thinning medial axis. Finally, we verify the method using the regular models and describe a random network micromodel quantitatively. The hybrid method not only accurately guarantees the medial axis located in the middle center but also keeps all the structure information of pore dead ends.