Image-Based Soil Particle Size and Shape Characterization.

Particle Size Distribution;Image Analysis;Wavelet Transformation;Watershed Segmentation;Civil and Environmental Engineering;Engineering;Civil Engineering

Ohm, Hyon-SohkZekkos, Dimitrios ;

University of Michigan

关键词: Particle Size Distribution;    Image Analysis;    Wavelet Transformation;    Watershed Segmentation;    Civil and Environmental Engineering;    Engineering;    Civil Engineering;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/97957/hyonsohk_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Two laboratory tests using image-based methods to determine particle size distribution of soil were developed. The Sediment Imaging or ;;Sedimaging” test determines size distribution of soil having particle diameters between 0.075 mm and 2 mm. A Translucent Segregation Table (TST) test determines size distribution of soil having particle diameter larger than 2 mm. Both tests produce particle size distributions that compare well with results by sieving. The Sedimaging test utilizes a statistical method based on wavelet transformation. The method requires images of relatively uniform particle sizes, thus sedimentation of a soil specimen through a column filled with water is implemented to segregate particles by size. An image of the sedimented soil is analyzed incrementally by overlapping 128 pixel by 128 pixel areas yielding thousands of values of a wavelet index (CA). The CA values are converted to particle sizes in units of pixels per particle diameter (PPD) through a previously established calibration curve. The calibration curve is an empirical fit to CA versus PPD data obtained from images of pre-sieved soils. The PPD is converted to sieve opening size using the known camera magnification. The TST test utilizes a deterministic method facilitated by watershed segmentation. The method requires thresholded images, thus a translucent plate and a backlight table are implemented to provide a bright and uniform grayscale contrast to the soil particles. Particles are introduced at the top of the inclined TST and are allowed to pass beneath the series of bridges having decreasing underpass heights. The bridges prevent small particles from hiding beneath large particles. After capturing an image of roughly segregated particles from above, touching particles in the image are segmented by the watershed segmentation. For each segmented particle, the largest and intermediate dimensions are computed by fitting an ellipse to the particle. The smallest dimension, which is not shown in the TST image, is estimated from the average of two bounding bridge heights between which each particle comes to rest on the TST. To correct the intermediate dimension to sieve opening size, a correction factor is derived based on the ratio between the intermediate and smallest dimensions.