The use of the fractal dimension in images for plant recognition

Fractal information from images can be used as a tool for plant recognition. Depending on the application plant images at various scales can be used. For the distinction between certain weeds in row crops images with a low number of plants (excluding total canopy cover) are used. To determine the fractal dimension, links with the Fourier transform are exploited. However, scatter in the power spectrum is reduced by integration in the spatial frequency domain over band-limited spatial filters. Using this approach the fractal dimension is determined for whole images and subsets of these images. The usefulness of the resulting fractal dimension information as a tool for image segmentation and crop versus weed recognition are discussed.