Weed and crop species classification using computer vision and deep learning technologies in greenhouse conditions

• DOI: 10.1016/j.jafr.2022.100325
• Publication Date: 2022-06-21T16:21:43Z
• Authors: Sunil G C; Yu Zhang; Cengiz Koparan; Mohammed Raju Ahmed; Kirk Howatt; Xin Sun
• Keywords: RGB color model

Site-specific weed management in Precision Agriculture is becoming a popular topic among researchers and farmers. The objective of this study was to classify weeds crop species using RGB image texture features with the comparison Support Vector Machine (SVM) classification model deep learning-based visual group geometry 16 (VGG16) models. A total 3792 images samples were captured from greenhouse, including 2271 1521 images. ReliefF feature selection algorithm applied select most important for prediction SVM VGG16 learning classifiers used four (horseweed, kochia, ragweed, waterhemp) six (black bean, canola, corn, flax, soybean, sugar beets). Accuracy, f1-score, kappa score metrics evaluate performance data reliability. had outperformed all classifiers. results showed that average f1-scores classifier obtained between 93% 97.5%. f1-score value 100% corn class Weeds-Corn classifier, which seems outstanding production system. This shows promising identification site-specific precision agriculture.