Online chicken carcass volume estimation using depth imaging and 3-D reconstruction.

Area of Science:

• Agricultural Engineering
• Computer Vision
• Machine Learning

Background:

• Variability in slaughtered chicken size poses challenges for poultry industry standardization.
• Current grading methods may lack precision and efficiency in assessing carcass size.

Purpose of the Study:

• To develop and validate a novel approach for chicken carcass grading using volume as a key metric.
• To implement real-time data capture and 3-D reconstruction for accurate volume estimation.

Main Methods:

• Utilized Kinect v2 depth imaging for real-time data capture of moving chicken carcasses.
• Employed 3-D reconstruction from point clouds and surface integration for volume calculation.
• Extracted 2-D and 3-D features for machine learning model input, specifically evaluating a bagged tree model.

Main Results:

• The bagged tree model achieved high accuracy with an R² of 0.9988, RMSE of 5.335, and ARE of 2.125%.
• The method demonstrated remarkable efficiency, with an average processing time of less than 1.6 seconds per carcass.
• Validated the effectiveness of depth imaging, 3-D reconstruction, and machine learning for precise chicken carcass volume estimation.

Conclusions:

• The novel volume-based grading approach offers a significant improvement in accuracy and efficiency over existing methods.
• This technology provides a reliable solution for automated chicken carcass grading, addressing a critical industry need.
• Depth imaging and machine learning are highly applicable for developing comprehensive and efficient poultry grading systems.