A novel approach for weed type classification based on shape descriptors and a fuzzy decision-making method

Area of Science:

• Agricultural Science
• Computer Vision
• Botany

Background:

• Effective weed management requires differentiating between monocot (grasses) and dicot (broad-leaved) weeds for targeted herbicide application.
• Selective herbicide use increases efficiency compared to broadcast application, reducing costs and environmental impact.

Purpose of the Study:

• To develop and validate a novel methodology for accurate weed species discrimination.
• To enable selective weed control strategies in agricultural settings.

Main Methods:

• Utilized shape descriptors, including seven Hu moments and six geometric features, to characterize weed species.
• Employed RGB camera imagery from outdoor field conditions with varying lighting.
• Adapted four decision-making methods using selected shape descriptors as attributes for classification.

Main Results:

• Achieved a high success rate in discriminating between grass and broad-leaved weed species.
• Demonstrated the effectiveness of shape descriptors in characterizing weeds from real-world field images.

Conclusions:

• The proposed methodology offers a robust and accurate approach for weed species discrimination.
• This technique supports precision agriculture by enabling targeted weed management strategies.