Weed detection system using support vector machine

In the oil palm plantation in Malaysia, typically blanket spraying is applied to a whole field without regard to the species and location of the weeds in the field. This practice is uneconomical since some areas where no or few weeds exist will receive just as much herbicides as those areas with high densities of weed infestation. Many of these chemical are soil-applied herbicides which easily absorb to ground water and surface water supplies. To control the weed grow and to solve the problems, automatic weed detection system need to be employed. This paper presents the results of automatic classification of broad and narrow weed using feature vector extracted using a combination of Gabor filter and FFT, and classifier using the support vector machine (SVM) Results obtained revealed that the proposed technique results in higher classification accuracy compared to other techniques.