Weed Mapping Using Digital Images

This study proposes a system to determine the weed coverage percentage in no-till field of common bean using image processing, acquired in a regular grid pattern. Images are initially converted in an excess green index (ExG) image, then segmented by threshold to indicate the vegetation areas. Since crop lines appear in regular lines, and weed in a more scattered pattern, this feature can be used to discriminate crop regions. This is done by estimating local orientation using an average squared gradient method. The crop areas detection is enhanced by filtering the ExG image using Gabor filters. Using the estimated vegetation and crop areas, a weed percentage value was assigned to each image. Since these images were georeferenced, it was possible to construct weed coverage maps by means of a linear interpolation method. The validation was performed by comparing interpolated values of weed coverage generated by the automatic system with random georeferenced higher resolution images. Comparison between the reference and the automatically generated maps reveals the capability to coherently distinguish high from low weed infestation in similar regions. keywords: machine vision, gabor filter, directional field, weed mapping

[1]  M. Costa,et al.  Action of two herbicides on the microbial activity of soil cultivated with common bean (Phaseolus vulgaris) in conventional-till and no-till systems , 2006 .

[2]  Daniel Marçal de Queiroz,et al.  Weed mapping using a machine vision system , 2012 .

[3]  Alberto Tellaeche,et al.  A new vision-based approach to differential spraying in precision agriculture , 2008 .

[4]  E. Rech,et al.  Transgenic Dry Bean Tolerant to the Herbicide Glufosinate Ammonium , 2002 .

[5]  X Yang,et al.  Evaluation of biofilm image thresholding methods. , 2001, Water research.

[6]  Gilles Rabatel,et al.  AE—Automation and Emerging Technologies: Weed Leaf Image Segmentation by Deformable Templates , 2001 .

[7]  Sabih H. Gerez,et al.  Systematic Methods for the Computation of the Directional Fields and Singular Points of Fingerprints , 2002, IEEE Trans. Pattern Anal. Mach. Intell..

[8]  J. Carpenter,et al.  Herbicide Use on Roundup Ready Crops , 2000, Science.

[9]  G. Meyer,et al.  Color indices for weed identification under various soil, residue, and lighting conditions , 1994 .

[10]  Roland Gerhards,et al.  The Economic Impact of Site-Specific Weed Control , 2003, Precision Agriculture.

[11]  L. Galli-Terasawa,et al.  Genetic diversity of rhizobia in a Brazilian oxisol nodulating Mesoamerican and Andean genotypes of common bean (Phaseolus vulgaris L.) , 2011 .

[12]  Scott D. Noble,et al.  Site-specific weed management: sensing requirements— what do we need to see? , 2005, Weed Science.

[13]  S. Christensen,et al.  Colour and shape analysis techniques for weed detection in cereal fields , 2000 .

[14]  R. Sichieri.  Dietary patterns and their associations with obesity in the Brazilian city of Rio de Janeiro. , 2002, Obesity research.

[15]  Anil K. Jain,et al.  Fingerprint Image Enhancement: Algorithm and Performance Evaluation , 1998, IEEE Trans. Pattern Anal. Mach. Intell..

[16]  Anil K. Jain,et al.  Unsupervised texture segmentation using Gabor filters , 1990, 1990 IEEE International Conference on Systems, Man, and Cybernetics Conference Proceedings.

[17]  J. Daugman.  Uncertainty relation for resolution in space, spatial frequency, and orientation optimized by two-dimensional visual cortical filters. , 1985, Journal of the Optical Society of America. A, Optics and image science.