| Abstract: | Automatic distinction of weed species using form parameters by means of digital image processing
In former investigations the possibility to distinguish weed from soil background by computerized quantitative image analysis has been reported. In the present studies image analysis technique are applied to identify different weed species. From six typical winter weed species including various growth stages, 300 color slides were taken at the research station in Bonn on the 4th of February, 1988. From each slide a number of form parameters were measured automatically out of the digitized shape of each weed plant. Form parameters including area, perimeter, minimum and maximum diameter of the plants are taken from the digitized shape of each weed as well as figures representing the relative proportions of these parameters. For automatic discrimination up to 9 of the parameters were selected to compute five so-called canonical variables. The values of these variables are typical for each weed species and can be interpreted as their "finger prints". The average rate of correct identification was 82.3 %. The scientific and practical implications of the results are discussed. |
