Robust erosion measurement in agricultural fields by colour image processing and image measurement

• Published in: European journal of soil science Vol. 64; no. 1; pp. 80 - 91
• Main Authors: Filin, S., Goldshleger, N., Abergel, S., Arav, R.
• Format: Journal Article
• Language: English
• Published: Oxford, UK Blackwell Publishing Ltd 01.02.2013; Blackwell
• Subjects: Agronomy. Soil science and plant productions; Biological and medical sciences; Color; Colour; Earth sciences; Earth, ocean, space; Erosion; Erosion mechanisms; Exact sciences and technology; Fundamental and applied biological sciences. Psychology; Image processing; Mathematical models; Soil (material); Soil science; Soils; Surficial geology; image processing; erosion; Cultivated field; Soil science; Earth science; Color image
• ISSN: 1351-0754; 1365-2389
• DOI: 10.1111/ejss.12010

Soil erosion is defined as a displacement of solid particles originating from soil, rock and other sediments by the forces of wind and water. Erosion affects soil micro‐topography and surface roughness and may lead to clogging of drainage systems, flooding and destruction of the upper part of the soil structure, thereby causing damage to cultivated land. Thus, quantification of changes in the amount and rate of soil erosion is vital for agricultural planning as well as for setting up different soil‐conservation systems. This paper describes an image‐based scheme for quantifying amounts of erosion and estimating the change in volume of raised beds resulting from irrigation or rainstorms in cultivated fields. Unlike existing methods that are labour‐intensive and error‐prone, the proposed approach is autonomous, accurate and can be applied with low‐cost digital cameras. Emphasis is placed on robustness to actual field conditions, including variations in illumination and shading and imaging from general vantage points. To accommodate varying field conditions, we propose the use of colour image‐processing, illumination‐invariant algorithms and linear photogrammetric models. Linear models eliminate the complex task of approximating camera‐position parameters, and introduction of colour‐driven image processing facilitates a computationally robust algorithm for volume estimation. Thus, the end‐product is a model that can be used by those who are unskilled in photogrammetry and image processing, with little or no guidance. Experiments show that estimates of volume and erosion reach accuracy levels of about 0.07 m3 ha−1.