An Estimation of the particle-size Distribution in gravel bed river Using Image Processing

Distribution pattern of the river bed particles grading creates important issues in investigation of the hydraulic, geomorphological and ecological behavior of the river.. For example, surface grain-size variability is crucial for illustrating sediment transport (Hoey and Ferguson, 1994; Russ, 1999; Joyce et al., 2001). Particle size characteristics that are dependent on particle size distribution are estimated in different ways, such as sieving method, sampling techniques in the field, photographic print method, and the other methods that have been suggested so far (Aberle and Nikora, 2006). The most prevalent method is the sieving method that obtains particle size distribution curve using cumulative weight of passing aggregation,.. It is obvious that, measurement of the particle size distribution based on field methods are time consuming, overwhelming and non-economic, so developing a fast and accurate method for measuring the particle size distribution of the river bed has a significant effect on civil and environmental engineering. Nowadays, this process is possible using image processing methods to automatically extract particle size using digital images of river bed. Various methods have been reported which aim to provide robust and automated estimates of grain size from images, falling under two broad categories classified by (Buscombe et al., 2010) as, respectively, ‘geometrical’ and ‘statistical’. Both techniques require imagery where the smallest grains are resolved by at least a few pixels. Statistical methods characterize grain size using a measure sensitive to image texture. These approaches have used autocorrelation (Warrick et al., 2009), semi variance or one of the several other methods, including fractals (Buscombe, 2013) and grey-level co-occurrence matrices. Geometrical methods use image processing techniques (principally, threshold and segmentation) to isolate and measure the visible axes (or portions of whole axes) of each individual grain (e.g. Graham et al., 2005; Chang &Chung, 2012). This research showed that new methods of image processing have an adequate potential to replace with previous traditional methods. Also, the percentage of human bias in methods such as field sampling or Sticky layer is very high for sampling of the Armor layer of the river bed, while the image processing method has enough power to take all the details and analysis of the data and it has a desirable accuracy compared to traditional methods. This resear