RELATIONSHIP ANALYSIS BETWEEN TEXTURE INFORMATION OF GRAVEL IMAGES AND HYDRAULIC CONDUCTIVITY

Abstract

 We study a method to extract texture information from gravel images and estimate hydraulic conductivity from texture information as a basic experiment. As texture information, we use 14 features of Heralic et al. calculated from the co-occurrende matrix. These features are invariant with respect to the parallel translation, scaling, and rotation in 45 degree increments of the image. Correlation analysis and multiple regression analysis are performed to analyze the relationship between texture information and hydraulic conductivity. In order to verify the effectiveness of the proposed method, gravel samples were prepared using river sand collected from the Urakami River, Nagasaki City, and experiments using its surface images and saturated hydraulic conductivity were performed. If the proposed method can be established, hydraulic conductivity can be obtained more easily and inexpensively than the conventional permeability tests.