The discrimination of crops, weeds and soil by optical reflectance was tested for broccoli, cabbage and leek crops from several different farms. The effects of crop varieties (in particular cabbage), weed population and type of soil did not affect the discrimination accuracies. No effects on crop/weed/soil discrimination were noted from the analysis of stressed crop samples (nitrogen concentration and water stress). Crop growth stage also had no influence on crop/weed/soil discrimination. Crop/weed/soil segregation into their respective groups required the use of classifiers. Two sets of spectral measurements, each comprising three wavelengths, were selected from all the data analyses providing the best overall accuracies. Discriminant analysis provided classification functions which differed greatly between farms. Neural networks provided the final algorithm relating the wavelength sets obtained by discriminant analysis. The use of broadband spectral range for discriminating between crop, weed and soil was also considered. This algorithm based on the discriminant integration index, also uses the discriminant analysis results to obtain the spectral range, but requires only one filter for accurate plant recognition. High discrimination accuracies are achieved with both algorithms and the broadband filter system shows potential for simplication without loss of performance in distinguishing between crops, weeds and soil.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:651938 |
| Date | January 1996 |
| Creators | Hahn, Federico |
| Publisher | University of Edinburgh |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://hdl.handle.net/1842/13977 |
Page generated in 0.002 seconds