Department of Geography
Loughborough University
Tel: +44 (0) 1509 22 2794
Fax: +44 (0) 1509 22 3930
Loughborough University

Automated Grain Sizing for Fluvial Gravels


You are here: Resources Gateway home > AGS > Background

Background

Why measure river-bed grain size?

The nature of the sediment in a river bed is a critical control on sediment transport rates, water flow velocity and aquatic habitat quality. Accurate characterisation of grain size is therefore essential in, for example, assessment of likely erosion around bridge footings, modelling patterns of flooding, and evaluating the impact of gravel extraction on economically important fish spawning grounds.

Why develop a new method of grain-size measurement?

Conventional grain-size measurement techniques are time-consuming and costly, requiring the manual collection, measurement and recording of hundreds of individual grains. The removal of grains for measurement damages or destroys the surface being studied, making conventional approaches unsuitable for ecological monitoring programmes. Furthermore, grain size is highly spatially variable and conventional appraches are unable to characterise this variation adequately, leaving potentially significant differences unrecognised.

These problems mean that the development of sampling techniques that achieve satisfactory characterisation of grain size whilst simultaneously reducing the time spent in the field and laboroatory is highly desirable. Several researchers have used emulsion-based photographic data capture to reduce field time, but subsequent analysis of the photographs can be extremely time-consuming.

These limtations can be overcome using automated methods of extracting information from images. Recent years have seen several groups attempt to characterise automatically from digital images the surface grain-size distribution of fluvial gravels exposed above the water surface. Although the results obtained have been encouraging, the test data sets used were small and/or limited to individual rivers and the transferability of such approaches to a range of lithotypes, grain shapes, packing configurations and sizes is unproven.

Principles

We have developed a new method of extracting grain-size information from digital images of exposed gravel surfaces. In developing this method we have employed three fundamental princples:

We have adopted a two-stage approach to developing and testing our automated grain sizing method. The first stage involved the development and testing of algorithms for the identification of objects (grains) in the images. The basis of this testing was against manually-digitised grain boundaries. A key objective for this stage was to identify an algorithm that perfomed well across images derived from a range of lithological provinces and not be tailored to any particular river or set of sedimentary circumstances. This stage is described in Graham, Reid and Rice (2005), and the object-detection algorithm section of the website augments the material in this paper.

The second stage applied the selected object-detection algorithm as part of an automated grain sizing method that corrects for factors such as lens and perpective distortions and evaluated the performance of the method against manually collected and graded sediment. This stage is described in Graham, Rice and Reid (2005), and the AGS procedure evaluation section of the website augments the material in this paper.

Both of these papers are available from the publications section of this website.


Website Maintained by: D.J.Graham@lboro.ac.uk - LU Home - Accessibility - © Loughborough University