Related Experiment Video
Updated: Jul 4, 2025
10:52
Protocols for Robust Herbicide Resistance Testing in Different Weed Species
Published on: July 2, 2015
14.6K
Weed detection in soybean fields using improved YOLOv7 and evaluating herbicide reduction efficacy
Jinyang Li1, Wei Zhang1,2, Hong Zhou1
1College of Engineering, Heilongjiang Bayi Agricultural University, Daqing, China.
Frontiers in Plant Science
|January 26, 2024
Summary
Sustainable agriculture requires herbicide reduction. A new YOLOv7-FWeed model using unmanned aerial vehicles (UAVs) accurately detects weeds in soybean fields, enabling reduced herbicide use without harming crops.
Area of Science:
- Agricultural Science
- Computer Vision
- Sustainable Agriculture
Background:
- Increasing environmental awareness drives the need for sustainable agriculture and reduced herbicide use.
- Current weed detection methods in soybean fields lack accuracy and efficiency, often relying on manual labor and struggling in complex environments.
- Accurate weed detection is crucial for evaluating the effectiveness of reduced herbicide application strategies.
Purpose of the Study:
- To develop an improved weed detection model for soybean fields to support herbicide reduction.
- To evaluate the effectiveness of reduced herbicide application levels on weed control and soybean growth.
- To provide an intelligent solution for efficient weed management in soybean cultivation.
Main Methods:
- Conducted weeding experiments in soybean fields with four levels of reduced herbicide application.
- Utilized an unmanned aerial vehicle (UAV) for image acquisition.
- Developed and implemented a novel weed detection model, YOLOv7-FWeed, incorporating F-ReLU activation and a MaxPool multihead self-attention (M-MHSA) module, based on YOLOv7.
Main Results:
- The YOLOv7-FWeed model demonstrated superior performance over YOLOv7 and YOLOv7-enhanced, achieving a precision of 0.9496, recall of 0.9125, F1-score of 0.9307, and mAP of 0.9662.
- An electrostatic spraying + 10% herbicide reduction level was identified as effective for weeding in soybean fields.
- Continuous monitoring confirmed that reduced herbicide application effectively controlled weed growth without negatively impacting soybean leaf area or dry matter weight.
Conclusions:
- The YOLOv7-FWeed model offers a highly accurate and efficient solution for weed detection in soybean fields.
- Reduced herbicide application, specifically electrostatic spraying with a 10% reduction, is a viable strategy for sustainable soybean farming.
- This research promotes intelligent weed management, supporting herbicide reduction and guiding efficient application techniques.
