Resumen de Multispectral classification of grass weeds and wheat (Triticum durum) crop using linear and nonparametric functional discriminant analysis, and neural networks

Francisca López Granados, José Manuel Peña Barragán, Montserrat Jurado Expósito, Mario Francisco-Fernández , Ricardo Cao Abad , Amparo Alonso Betanzos

• Field research was conducted to determine the potential of multispectral classi¯cation of late-season grass weeds in wheat. Several classi¯cation tech- niques have been used to discriminate di®erences in re°ectance between wheat, and wild oat, canarygrass, ryegrass and rabbit foot in the 400-900 nm spectrum, and to evaluate the accuracy performance for a spectral signature classi¯cation into the plant species or group to which it belongs. Fisher linear discriminant analysis, nonparametric functional discriminant analysis and several neural net- works have been applied, either with a preliminary principal component analysis or not and in di®erent scenarios. Fisher linear discriminant analysis, feedforward neural networks and one-layer neural network, all of them with a principal com- ponent analysis, showed classi¯cation percentages of between 90% and 100 %.

  Generally, a preliminary computation of the most relevant principal components considerably improves the correct classi¯cation percentage. These results are promising because wild oat and ryegrass, two of the most problematic, clearly patchy and expensive to control weeds in wheat, could be successfully discrim- inated from wheat in the 400-900 nm range. Our results suggest that mapping grass weed patches in wheat could be feasible with image analysis working for real-time and high-resolution satellite imagery acquired in mid-May under our conditions.