Detection of Weeds In Unstructured Wheat Field Using Image Processing And Machine Learning

EasyChair Preprint 3798

Abstract

Weed distribution levels range between low and excessive densities. Two computer vision-based algorithms are presented in this paper to identify widespread weeds in wheat fields under natural field conditions. First algorithm explores weeds by image processing rules. Algorithm used color to differentiate flowers from soil. While texture analysis strategies are used to distinguish weeds from crops than in the second step multi class linear kernel SVM used for classification of the images whether it is a wheat field or weed based on the weed thicknesses which is shown in images. Back propagation and RBF kernel SVM used for comparison between results. On the basis of execution time and accuracy back propagation neural network outperform rather than multi-class linear kernel SVM shows better result.

Keyphrases: Operations, Weeds, computer vision, image processing, morphological

@booklet{EasyChair:3798,
  author    = {Muhammad Umair Khan},
  title     = {Detection of Weeds In Unstructured Wheat Field Using Image Processing And Machine Learning},
  howpublished = {EasyChair Preprint 3798},
  year      = {EasyChair, 2020}}