The Posture Detection Method of Caged Chickens Based on Computer Vision.

Area of Science:

• Agricultural Technology
• Computer Vision
• Animal Science

Background:

• Manual monitoring of caged chickens is labor-intensive and time-consuming.
• Automated systems are needed to improve efficiency and welfare in poultry farming.
• Posture analysis provides insights into animal health and behavior.

Purpose of the Study:

• To develop an automated computer vision-based method for detecting the postures of caged chickens.
• To reduce the manual labor involved in monitoring chicken status in farming.
• To provide a non-invasive tool for analyzing chicken behavior and welfare.

Main Methods:

• Image correction techniques were applied to standardize image orientation.
• Variance and Speeded-Up Robust Features (SURF) methods were used to locate feeding troughs and define key areas.
• A depth camera was utilized to extract 3D information and segment chickens within the key areas.
• Constraint conditions were applied to screen and classify chicken postures (standing/lying).

Main Results:

• The algorithm achieved high precision and recall rates for posture detection in both white (97.80% precision, 80.18% recall) and jute chickens (79.52% precision, 81.07% recall).
• The system demonstrated efficient processing, running at ten frames per second on an i5-8500 CPU.
• The method proved effective in extracting chickens and analyzing their postures from complex cage environments.

Conclusions:

• The proposed computer vision method offers an accurate and efficient non-invasive approach for monitoring caged chicken postures.
• This technology has the potential to significantly improve poultry management by automating status checks.
• The findings support the advancement of precision poultry farming and animal welfare research.