Estimation of riverbed grain-size distribution using image-processing techniques

Quantification of the grain size distribution of fluvial gravels remains an important and challenging issue in the study of river behavior. It is desirable for sampling techniques to achieve accurate estimation of grain size distribution, while simultaneously reducing the time spent. Recent advances in image analysis techniques have facilitated automated grain identification and measurement within digital images. In this study, an image-processing method fusing feedback pulse couple neural network and multilevel thresholding, the I-FM method, is proposed for automatic extraction of grain-size distribution based on digital photographs taken from a river-bed. A decisive image-merging algorithm is also developed for improving the quality of image segmentation in grain-size measurements. The experiments were conducted in both lab and field, and the proposed method was compared with traditional image processing methods. The proposed I-FM produces much more satisfactory results in estimating the amount of gravel and the percentiles of grain-size distribution in comparison with other image processing methods and manual sieving methods. It demonstrates the I-FM method is an efficient method for precisely measuring the grain-size distribution of river-bed material.