Related Experiment Video
Updated: Jun 27, 2026
12:08
The Use of Traditional Fear Tests to Evaluate Different Emotional Circuits in Cattle
Published on: April 22, 2020
Beef Cattle Behavior Recognition Based on Nighttime Farm Videos via Spatio-Temporal Enhancement and Dynamic Fusion
Yamin Han1, Zhenyu Zhang1, Wenchao Zhang1
1College of Information Engineering, Northwest A&F University, Yangling 712100, China.
Animals : an Open Access Journal From MDPI
|June 26, 2026
Summary
This study introduces a new AI method for recognizing beef cattle behavior in dark farm conditions. The novel approach enhances dark videos and fuses features, improving accuracy for nighttime livestock monitoring.
Area of Science:
- Agricultural Science
- Computer Vision
- Animal Behavior
Background:
- Beef cattle behavior analysis is crucial for health monitoring.
- Current deep learning models struggle with low-light conditions in real farm settings.
- Robust nighttime behavior recognition is needed for precision farming.
Purpose of the Study:
- To develop a reliable method for beef cattle behavior recognition in dark environments.
- To create a new dataset of nighttime beef cattle behaviors.
- To improve the application of AI in livestock monitoring.
Main Methods:
- Constructed the "Dark Beef Cattle Actions" dataset with 1097 nighttime video clips.
- Proposed a novel neural network with spatio-temporal dark enhancement and dynamic fusion.
- Utilized a joint loss function for optimizing dark enhancement and action recognition.
Main Results:
- Achieved 88.47% precision, 80.18% recall, 83.80% accuracy, and 84.12% F1-score on the Dark Beef Cattle Actions dataset.
- The proposed method demonstrated competitive performance against state-of-the-art techniques.
- Successfully recognized 6 key beef cattle behaviors in nighttime conditions.
Conclusions:
- The developed AI method effectively recognizes beef cattle behavior in low-light conditions.
- This research supports intelligent livestock monitoring and precision farming advancements.
- The "Dark Beef Cattle Actions" dataset is a valuable resource for future research.