Discrimination of sunflower, weed and soil by artificial neural networks


Kavdir I.

COMPUTERS AND ELECTRONICS IN AGRICULTURE, cilt.44, ss.153-160, 2004 (SCI İndekslerine Giren Dergi) identifier identifier

  • Cilt numarası: 44 Konu: 2
  • Basım Tarihi: 2004
  • Doi Numarası: 10.1016/j.compag.2004.03.006
  • Dergi Adı: COMPUTERS AND ELECTRONICS IN AGRICULTURE
  • Sayfa Sayıları: ss.153-160

Özet

Selective application of herbicide to weeds at an early stage in crop growth is an important aspect of site-specific management of field crops, both economically and environmentally. This paper describes the application of a neural network classifier to differentiate between 2 and 3 weeks old sunflower plants and common cocklebur weeds of similar size, shape and colour. Colour images were obtained by a digital camera, in natural sunlight. A specific objective was to minimise the subsequent image processing operations needed to enhance the images and to extract the features needed by a back propagation neural network classifier. Neural network structures with different numbers of hidden layers and neurons in them were tested to find the optimal classifier. The maximum number of correctly recognised images in distinguishing weeds from sunflower plants was 71 (out of 86), while it was 82 and 74 in separating sunflower and weed images from bare soil images, respectively. (C) 2004 Elsevier B.V. All rights reserved.