Related Experiment Video
Updated: Dec 23, 2025
08:47
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
2.0K
Weed Classification for Site-Specific Weed Management Using an Automated Stereo Computer-Vision Machine-Learning
Mojtaba Dadashzadeh1, Yousef Abbaspour-Gilandeh1, Tarahom Mesri-Gundoshmian1
1Department of Biosystems Engineering, College of Agriculture and Natural Resources, University of Mohaghegh Ardabili, Ardabil 56199-11367, Iran.
Plants (Basel, Switzerland)
|May 1, 2020
Summary
This study introduces a stereo vision system using artificial neural networks (ANNs) and metaheuristic algorithms to accurately distinguish rice from weeds. The developed system achieved high accuracy, improving eco-friendly weed management in rice cultivation.
Area of Science:
- Agricultural Engineering
- Computer Vision
- Artificial Intelligence
Background:
- Site-specific weed management in dense crops like rice is challenging.
- Selective herbicide application requires accurate plant and weed discrimination.
- Current methods often lack the precision needed for effective, eco-friendly weed control.
Purpose of the Study:
- To develop a stereo vision system for distinguishing rice plants from weeds.
- To further discriminate between two specific types of weeds within a rice field.
- To enhance weed management strategies through advanced image analysis and classification.
Main Methods:
- Recorded stereo videos in a rice field and processed frames.
- Extracted green plants from the background after pre-processing and segmentation.
- Utilized artificial neural networks (ANNs) optimized with particle swarm optimization (PSO) and the bee algorithm (BA) for feature selection and classification.
- Extracted 302 color, shape, and texture features for discrimination.
Main Results:
- The proposed ANN-BA classifier achieved high accuracies (88.74% right, 87.96% left channels).
- Accuracies improved to 92.02% (arithmetic mean) and 90.7% (geometric mean).
- The ANN-BA classifier significantly outperformed the K-nearest neighbors (KNN) classifier, which had lower overall accuracy.
Conclusions:
- The developed stereo vision system effectively distinguishes rice from weeds with high accuracy.
- The integration of ANNs and metaheuristic algorithms offers a promising solution for site-specific weed management.
- This technology can contribute to more sustainable and eco-friendly agricultural practices in rice cultivation.
