Automated Crop and Weed Monitoring in Widely Spaced Cereals

Abstract

An approach is described for automatic assessment of crop and weed area in images of widely spaced (0.25 m) cereal crops, captured from a tractor mounted camera. A form of vegetative index, which is invariant over the range of natural daylight illumination, was computed from the red, green and blue channels of a conventional CCD camera. The transformed image can be segmented into soil and vegetative components using a single fixed threshold. A previously reported algorithm was applied to robustly locate the crop rows. Assessment zones were automatically positioned; for crop growth directly over the crop rows, and for weed growth between the rows. The proportion of crop and weed pixels counted was compared with a manual assessment of area density on the basis of high resolution plan view photographs of the same area; this was performed for views with a range of crop and weed levels. The correlation of the manual and automatic measures was examined, and used to obtain a calibration for the automatic approach. The results of mapping of a small field, at two times, are presented. The results of the automated mapping appear to be consistent with manual assessment.

Automatic segmentation of relevant textures in agricultural images

“…And the vegetative index (VEG) described in Hague et al (2006), which was designed to cope with the variability of natural daylight illumination. All these approaches need to fix a threshold for final segmentation.…”

Automatic expert system based on images for accuracy crop row detection in maize fields

“…Vegetation indices to be combined are computed as follows Guijarro et al (2011), Excess Green (Woebbecke et al, 1995;Ribeiro et al, 2005) Color index of vegetation extraction (Kataoka, Kaneko, Okamoto, & Hata, 2003) Vegetativen (Hague et al, 2006) Neto, 2004) where excess red is computed as follows (Meyer, Hindman, & Lakshmi 1998): ExR = 1.4r-g. According to Guijarro et al (2011) the above four indices are combined to obtain the resulting value COM as follows,…”

Effective segmentation of green vegetation for resource-constrained real-time applications

“…Significant research has been undertaken on the segmentation of green vegetation for crop row line tracking and identification of single plants (crops and weeds) for applications such as precision spraying. Visible, spectral index-based methods, which have been proposed to segment green vegetation under variable illumination conditions include Excess Green (ExG) (Woebbecke et al, 1995), Color Index of Vegetation Extraction (CIVE) (Katoka et al, 2003), Excess Green minus Excess Red (ExGExR) and Vegetative index (VEG) (Hague et al, 2006). A critical step is required to select the threshold value to binarize the near-binary image resulting from all the above spectral index-based methods.…”