A generalized model for accurate wheat spike detection and counting in complex scenarios

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Accurate wheat spike detection and counting are essential for crop yield estimation and breeding.
• Challenges exist in complex field environments due to visual similarities and occlusions.

Purpose of the Study:

• To develop an effective model, RIA-SpikeNet, for detecting and counting wheat spikes in challenging field conditions.
• To improve the accuracy and robustness of wheat spike analysis.

Main Methods:

• Introduction of an Implicit Decoupling Detection Head for better discrimination of similar spikes.
• Utilization of Asymmetric Loss to enhance learning for positive and hard samples.
• Modification of the backbone network with reparameterization and larger convolutional kernels to improve feature extraction.

Main Results:

• RIA-SpikeNet achieved a competitive 81.54% mAP and 90.29% R².
• The model demonstrated superior performance compared to the state-of-the-art YOLOv8 detection model.
• Enhanced accuracy in detecting and counting wheat spikes in complex field scenarios.

Conclusions:

• RIA-SpikeNet offers an effective solution for wheat spike yield estimation in field settings.
• The model provides valuable support for wheat production and breeding efforts.
• The proposed techniques significantly improve wheat spike detection and counting accuracy.