EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Segmentation of weeds and crops using multispectral imaging and CRF-enhanced U-Net

Halil Mertkan Sahin, Tajul Miftahushudur, Bruce Grieve, Hujun Yin

Computers and Electronics in Agriculture · 2023-08 · 211 · 107956

DOI: 10.1016/j.compag.2023.107956 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Computers and Electronics in Agriculture, 2023-08. Verfasst von Halil Mertkan Sahin und 3 weiteren Autor:innen. Die Arbeit wurde bisher 78-mal zitiert (Crossref) und nennt 55 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Weeds are one of the most damaging invaders that can cause great losses of crop yields as they compete with crops for water, nutrients and sunlight. Early detection of weeds in the field is very critical for taking early actions such as applying possible treatments or herbicides, which are not applicable after a certain stage of crop growth. In this work, a deep learning method has been developed for weed detection and a sunflower dataset consisting of multispectral images (RGB and NIR) captured at various days and times was used for the study. The deep learning method, U-Net, was trained with images of the emergence state and the subsequent growth state and tested on images of crops in the last state where chemical treatments can be applied. The results of the U-Net were further enhanced by employing the conditional random fields to achieve better segmentation in terms of Intersection over Union (IoU). The proposed method using the Green + Filtered-NIR + NDVI channels as the input achieved the best mean IoU score of 0.883 on images of 512 x 512 pixels. In the same experiment, soil, crop, and weed pixels were correctly predicted with 0.990 IoU, 0.906 IoU, and 0.753 IoU scores, respectively. The results show that the chosen input and the proposed method present a valid approach for crop/weed segmentation.

Zitieren

APA

Sahin, H. M., Miftahushudur, T., Grieve, B., & Yin, H. (2023). Segmentation of weeds and crops using multispectral imaging and CRF-enhanced U-Net. Computers and Electronics in Agriculture, 211, 107956. https://doi.org/10.1016/j.compag.2023.107956

Vancouver

Sahin HM, Miftahushudur T, Grieve B, Yin H. Segmentation of weeds and crops using multispectral imaging and CRF-enhanced U-Net. Computers and Electronics in Agriculture. 2023;211:107956. doi:10.1016/j.compag.2023.107956

BibTeX
@article{Sahin2023,
  author = {Sahin, Halil Mertkan and Miftahushudur, Tajul and Grieve, Bruce and Yin, Hujun},
  title = {Segmentation of weeds and crops using multispectral imaging and CRF-enhanced U-Net},
  journal = {Computers and Electronics in Agriculture},
  year = {2023},
  volume = {211},
  pages = {107956},
  doi = {10.1016/j.compag.2023.107956}
}
Kostenlos starten →