01783nam a2200265 a 450000100080000000500110000800800410001910000210006024500920008126001950017330000130036850000260038152009250040765000190133265000150135165000110136665000130137765300230139065300160141365300190142965300180144865300170146670000170148370000170150010090262020-01-17       2004    bl uuuu     u00u1 u    #d1 aCAMARGO NETO, J.  aAdvances in color image segmentation of plants for weed control.h[electronic resource]  aIn: ASAE/CSAE ANNUAL INTERNATIONAL MEETING, 2004, Ottawa. Dynamic partnerships for an environmentally safe and healthy world. St. Joseph, MI: American Society of Agricultural Engineersc2004  ap. 1-16.  aPaper number: 043060.  aAbstract. A new and improved unsupervised plant segmentation index (ExGExR) is introduced. ExGRxR performance was compared with the traditional excess green segmentation method (ExG), using Otsu's method for automatic threshold. Segmentation accuracies were measured with a quality index, using a ground truth target plant mask, hand-generated for each image. Statistical results showed that there were significant differences in performance with bare soil backgrounds during the first week for pigweed and velvetleaf. For corn stalk residue, there were significant differences for all species during the first week and also significant differences during the second week for pigweed and velvetleaf. Using wheat straw residue background, there were significant differences for all species all three weeks. In general, the algorithm performed better than the ExG index for segmenting plants from soil-residues background.  aImage analysis  aNo-tillage  aPlanta  aResíduo  aAnálise de imagem  aColor index  aÍndice de cor  aSegmentação  aSegmentation1 aMEYER, G. E.1 aJONES, D. D.