Grain size analysis for coarse river beds using digital imagery processing

Quantification of the grain size distribution of river beds remains an issue of large importance for a wide range of river engineering problems. A new procedure was empirically derived from flume experiments performed at the Laboratory of Hydraulics, Hydrology and Glaciology (VAW), using digital imagery analyses. The proposed method represents a fast, easy and inexpensive alternative to determine the grain size distribution of the subsurface of coarse river beds.

In a first step, a photograph is taken of a dry bed section from an appropriate height. The ‘flying height’ of the camera should be chosen in order to provide that the image is of the correct scale with respect to the size of grains that are of interest. In order to conclude from a pixel-based evaluation to unit lengths, the scale of the image should be determined by placing a measuring-tape on the gravel surface at the sample site.

The purpose of the following image processing step is to convert the vertical photograph into a quasi-binary image (Fig. 1). The quasi-binary image is suitable for an automated extraction of grain size data, which is performed by the software “Image Tool 3.0” (freeware: http://ddsdx.uthscsa.edu/dig/itdesc.html). This programme scans the image to identify and analyse objects (Fig. 2). Besides other information, the longest and shortest axes of the recognised objects are measured. The analysis with automated object recognition resembles a point analysis of the surface layer. Thus, an appropriate conversion model was developed to determine the volumetric distribution of the subsurface layer.

The method was tested with independent flume and field data, resulting in promising approximations to the reference grading curves of the subsurface.