Elsevier

Biosystems Engineering

Volume 198, October 2020, Pages 63-77
Biosystems Engineering

Research Paper
Estimating soil aggregate size distribution from images using pattern spectra

https://doi.org/10.1016/j.biosystemseng.2020.07.012Get rights and content

Highlights

  • •

    Trained regression models able to predict soil aggregate size distribution from images.

  • •

    Identification of the most suitable hierarchical representation for attribute pattern spectra.

  • •

    Comparison to the classical structuring element spectra.

  • •

    Publishing the dataset with RGB images of soil samples at different scales, with their aggregate size distribution.

A method for quantifying aggregate size distribution from the images of soil samples is introduced. Knowledge of soil aggregate size distribution can help to inform soil management practices for the sustainable growth of crops. While current in-field approaches are mostly subjective, obtaining quantifiable results in a laboratory is labour- and time-intensive. Our goal is to develop an imaging technique for quantitative analysis of soil aggregate size distribution, which could provide the basis of a tool for rapid assessment of soil structure. The prediction accuracy of pattern spectra descriptors based on hierarchical representations from attribute morphology are analysed, as well as the impact of using images of different quality and scales. The method is able to handle greater sample complexity than the previous approaches, while working with smaller samples sizes that are easier to handle. The results show promise for size analysis of soils with larger structures, and minimal sample preparation, as typical of soil assessment in agriculture.

Section snippets

Introduction and motivation

Soil structure concerns the physical arrangement of a soil, which provides an environment to provide plants access to water, air and nutrients, and a suitable medium for root development (Bronick & Lal, 2005). Adequate soil structure is fundamental for the sustainable growth of crops, and can contribute to reducing the environmental impact of agriculture. Hence, robust and accurate methods to measure soil structure are important tools for informing soil management decisions.

Soil aggregates

Related work

Some of the earliest image processing techniques applied to estimation of aggregate size distribution focused on segmenting images of non-overlapping coarse aggregates (3 mm–63 mm) (Mora et al., 1998). An automated tool for measuring the grain size distribution of gravels from digital photographs was developed by Graham et al. (2005) and improved by Detert and Weitbrecht (2012) based on analysing a number of segmentation-based techniques for overlapping particles of coarse-grained sediments,

Methodology

In this section, we introduce the basic concepts of morphological image processing used in the implementation of the digital sieve from the highlighted part of the system in Fig. 1.

Dataset description

In this section, the dataset collected to study the application of image processing techniques to soil structure assessment is described. Soil samples were collected from arable soils using a spade to a depth of 200 mm, a method similar to initial soil extraction for other conventional in-field soil structural assessments (Ball et al., 2007). In order to minimise subsequent disturbance, blocks of soil were not sampled from direct areas of contact with the spade, and they were carefully placed

Experimental setup

This section presents the experiments designed to examine the ability to predict the soil size distribution from image pattern spectra based on different image hierarchies from attribute morphology, as well as SE morphology, and assess their potential for developing an imaging pipeline for performing quantitative soil analysis. All images have been loaded as greyscale for further processing, relying on the internal conventions of the libjpg codec for the conversion from colour images.

The bin

Results and discussion

As the error distributions were heavily tailed, the results were expressed in terms of median and median absolute deviation as e˜±MAD, where e˜=median(e) and MAD=median(|(ei−e˜)|). The results indicate that pattern spectra descriptors show promising ability in predicting the soil size distribution, with the best predictor resulting in the expected error in aggregate diameter of (1.1±10) mm when measuring both in terms of weight (Fig. 6) and volume (Fig. 7).

Such results suggest that the

Conclusions and future work

The suitability of pattern spectra for determining the soil aggregate size distribution from soil sample images has been confirmed by including more soil sample images and directly predicting the soil aggregate size distribution measured both in terms of mass and volume. Our experiments were designed to work with small sample sizes, and examined the performance of the descriptors under scale changes and the presence of visible background. The suitability of different component trees for soil

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (56)

  • M. Aitkenhead et al.

    Estimating soil properties with a mobile phone

  • D.E. Allen et al.

    Soil health indicators under climate change: A review of current knowledge

  • B.C. Ball et al.

    Field assessment of soil structural quality – a development of the Peerlkamp test

    Soil Use and Management

    (2007)
  • F. Bianconi et al.

    Grain-size assessment of fine and coarse aggregates through bipolar area morphology

    Machine Vision and Applications

    (2015)
  • N. Bonneel et al.

    Wasserstein barycentric coordinates: Histogram regression using optimal transport

    ACM Transactions on Graphics

    (2016)
  • P. Bosilj et al.

    Retrieval of remote sensing images with pattern spectra descriptors

    ISPRS International Journal of Geo-Information

    (2016)
  • P. Bosilj et al.

    Pattern spectra from different component trees for estimating soil size distribution

  • P. Bosilj et al.

    Partition and inclusion hierarchies of images: A comprehensive survey

    Journal of Imaging

    (2018)
  • D. Buscombe et al.

    Grain-size information from the statistical properties of digital images of sediment

    Sedimentology

    (2009)
  • G. Cavallaro et al.

    Automatic attribute profiles

    IEEE Transactions on Image Processing

    (2017)
  • Y. Chen et al.

    Gray-scale morphological granulometric texture classification

    Optical Engineering

    (1994)
  • J. Cousty et al.

    Hierarchical segmentations with graphs: Quasi-flat zones, minimum spanning trees, and saliency maps

    Journal of Mathematical Imaging and Vision

    (2018)
  • S. Czarnes et al.

    Root-and microbial-derived mucilages affect soil structure and water transport

    European Journal of Soil Science

    (2000)
  • M. Detert et al.

    Automatic object detection to analyze the geometry of gravel grains–a free stand-alone tool

  • E. Dougherty et al.

    Morphological image segmentation by local granulometric size distributions

    Journal of Electronic Imaging

    (1992)
  • A. Doulamis et al.

    Generalized multiscale connected operators with applications to granulometric image analysis

  • D. Graham et al.

    Automated sizing of coarse-grained sediments: Image-processing procedures

    Mathematical Geology

    (2005)
  • D. Graham et al.

    A transferable method for the automated grain sizing of river gravels

    Water Resources Research

    (2005)

    • Cited by (0)

    View full text
    © 2020 IAgrE. Published by Elsevier Ltd. All rights reserved.