Related Experiment Video
Updated: Aug 5, 2026
11:49
Cereal Crop Ear Counting in Field Conditions Using Zenithal RGB Images
Published on: February 2, 2019
Multiple Kernel Attention Network for Dense and Tiny Wheat Pest Detection in the Field Under Complex Background
Xiang Li1, Mingqiang Chen2, Lei Qian2
1School of Computer Science and Artificial Intelligence, Chaohu University, Hefei 238000, China.
Insects
|July 27, 2026
Summary
Accurate wheat pest detection is crucial for crop yield. A new Multiple Kernel Attention Network (MKA-Net) improves detection of dense and tiny pests, achieving 67.1% AP50.
Area of Science:
- Agricultural Science
- Computer Vision
- Machine Learning
Background:
- Pest outbreaks significantly impact wheat yield and quality.
- Accurate pest recognition is vital for early warning systems.
- Challenges include insufficient datasets, imbalanced samples, and dense, tiny pest distributions.
Purpose of the Study:
- To develop a high-quality wheat pest dataset.
- To address challenges in detecting dense and tiny pests.
- To improve the accuracy of wheat pest detection.
Main Methods:
- Collected a real-world dataset of wheat pest images over two years.
- Developed a cut-up data augmentation strategy for dense pest samples.
- Introduced the Multiple Kernel Attention Network (MKA-Net) integrating multi-scale pest features.
Main Results:
- Achieved a peak AP50 of 67.1% in wheat pest detection.
- Demonstrated a significant improvement of 6.9 points over the baseline.
- The MKA-Net effectively handles tiny body sizes and dense distributions.
Conclusions:
- The proposed method enhances wheat pest detection accuracy.
- This contributes to the prevention and control of wheat pests.
- Promotes the advancement of intelligent agriculture through improved pest management.