AUTOMATED-DETERMINATION OF THE DISTRIBUTION OF LOCAL VOID RATIO FROM DIGITAL IMAGES

The frequency distribution of local void ratio is believed to be an im portant parameter, in addition to the void ratio, for describing the m echanical behavior of granular materials. Oda proposed a method to det ermine experimentally the distribution of local void ratio from 2-D pl ane sections. To date. implementations of Oda's method have depended t o varying extents on operator judgment to form polygons by joining the centers of gravity of all particles that surround a void. Furthermore , the studies have involved a significant amount of manual work in mak ing the required measurements. This paper describes a fully automated implementation of the method. which uses high-level. image-processing techniques. The proposed method eliminates operator judgment and manua l work and makes the determination of the distribution of local void r atio from 2-D plane sections both repeatable and efficient. The method is illustrated with measurements performed on synthetic and real imag es. The importance of correcting the images to account for factors suc h as thickness of se mentation lines is demonstrated. Measurements tha t confirm the stability of the proposed polygon network generation pro cedure are also presented.