Digital image processing of mass-size distribution of soil structures in plough layer

Abstract: In many instances the basic structural units of plough layer are soil aggregates which are resulted from tillage operation and packed into layers to form the seedbed. Quantification of plough layer is limited to a few basic soil parameters, including cone index, bulk density and porosity. Soil aggregates are assessed with dry sieving. These parameters do not provide the detailed structural information of plough layer. Precision management of plough layer requires that soil structures be quantified with more parameters that are geometrically quantifiable. Quantitative method for tilled-layer soil structures was adopted; the digital images of soil structures were taken, the aggregate mass was calculated and the mass-size distributions of soil structures sampled from a plowed paddy field were studied. The tilt angle of camera was set to 30°, 45°, 60° and 90°, respectively, when taking the photos of soil structures. Soil structures were also measured manually with dry-sieving method for comparison. During the manual measurement of soil aggregates, a caliper was used and both the long and the short axes of the aggregates were measured. The dry-sieving used the nested sieves with the openings of 4, 8, 16, 32, 64 and 128 mm, respectively. Geometrical parameters of soil structures in each size range were calculated with an image-processing program developed in MatLab, including angularity, shape index and rectangle degree. Collected data for the mass-size distribution of the soil aggregates after plowing were also fitted respectively with 3 models, i.e. Weibull model, Rosin-Rammler model and Gaudin-Schuhmann model. It showed that, along with the increase of size range, both angularity and shape index increased, but rectangle degree decreased, meaning that different effects of mechanical operation were induced by different size ranges of soil structures, even though under the same plowing treatment. Detailed analysis on each size range and each tilting angle showed that photos taken with 60o tilting angle yielded the best fitting results compared with other tilting angles. The 60o tilting angle was the most suitable for camera when used for on-line soil structure monitoring. No significant difference (P>0.05) was observed between digital image processing and manual measurement, proving that the digital image processing was an accurate method to acquire mass-size distributions of soil structure. Compared with Weibull and Rosin-Bammler distribution, Gaudin-Schumann model provided the best fitting between aggregate mass and size, with the R2 of 0.98. Digital image processing discriminated soil structures in finer scales and provided a higher precision curve fitting for soil structures compared with dry-sieving method. The variation of the acquired results from dry-sieving was significant due to the large size ranges between adjacent sieve sizes. Unlike the limited methods for tilled-layer soil structure quantification, such as dry sieving, image-processing was capable of not only quantifying the geometrical parameters of soil structures, but also distinguishing and separating soil structures in finer scales, such as 5 mm size range or any other arbitrary scales. This fine scale distinction was helpful in providing more precise modeling on soil structures. The results prove that the image-processing is a powerful tool to calculate geometric parameters of soil structures and discriminate soil structural features in detail.