Optimal pixel resolution for sand particles size and shape analysis

In this paper, the influence of different scanning resolutions and the use of image analysis softwares for the image analysis, SigmaScan Pro, ImageJ and Matlab, on size and shape factors commonly used in particles characterization were investigated. In our previous paper [1] the characterization of polydisperse sand particles was performed, and the need had arisen to examine the impact of different scanning resolutions on shape factors in more detail. The particles analyzed were the polydisperse fractions of quartz filtration sand with sieve diameters in the intervals of 0.85–1.030mm, 1.406–1.600mm and 2.00–2.83mm. The particles size and shape were analyzed using the scanned image of about 150 particles in each fraction. In addition, the images of circles of similar dimensions were generated to serve as reference particles. The scanning resolutions used in this study were from 75 to 4800dpi, and grayscale thresholds were optimized for discrimination of the particles from the background. The obtained results indicate that the resolution (i.e. the pixel size) and the algorithms used in image analysis softwares have significant influence on the obtained shape factors. The most significant effect was observed in the calculated values of the particles perimeters, while the effect was less pronounced in the results obtained for the particles projected area. The analysis of images of generated circles confirmed that the resolution had the largest effect on particle perimeter. The use of the resolutions of 300–600dpi for the determination of particle shape and size can be recommended for particles of ~1mm and larger, because of reasonable results, low storage space and less time for image analysis. [Display omitted] •2D image analysis was used for the characterization of non-spherical particles.•Particle perimeter and circularity show the most dependence on the resolution.•Minimum particle/pixel size ratio for accurate particle characterization is 20.•Perimeter calculated in SigmaScan Pro is larger comparing with ImageJ and Matlab.•Analysis of circles as referent particles was performed for validation of the experiments.