Danish Institute of Agricultural Sciences, Department of Crop Protection,
Research Centre Flakkebjerg, DK-4200 Slagelse, PrebenK.Hansen@agrsci.dk
Weed species and density are the central parameters in estimation of the optimal herbicide dose. In a Site Specific Weed Management (SSWM) system, weed species and density are usually found by manually weed counting, which is an objective and simple method.
However, it is not realistic to use manually methods, if the farmer wants to conduct SSWM, because a large number of time consuming weed registrations are needed. An automatic weed monitoring system method will improve the possibilities for using SSWM in practical farming.
Several studies have described systems, that are able to detect weed species and density automatically, but none systems are commercial yet. Assuming, that such a system would be available, knowledge of the crop status (competitiveness) could probably be extracted from the system. This would improve the precision of the herbicide dose optimisation.
Because the decision of herbicide and dose takes place in the early growth stages, the relationship between early growth dynamics and the competitive ability has to be considered. At the Danish Institute of Agricultural Sciences, Department of Crop Protection a method for estimating the horizontal crop-leaf distribution in the early growth stages has been developed, and the relationship between crop competitiveness and early horizontal leaf distribution is under examination.
The method uses images acquired 1.5 m above the soil surface with a digital camera of the type Olympus C-1400 XL with a 32 mm 1:2.8-3.9 lens in “super high quality” (1280´1024 pixels). These images are converted from the RGB colour space to the HSI colour space. The image segmentation uses a multi-band thresholding technique in the HSI colour-space to separate green from non-green pixels.
A histogram of the spatial distribution of the green pixels shows a wave-formed curve. After fitting a sinusoidal function to the curve, the inter-rows of the crop are found where the curve has local minima. A four-parametric logistic peak function is then fitted to the distribution of green pixels from inter-row to inter-row.
The method and preliminary results of the relationship between crop competitiveness and horizontal leaf distribution are presented.