SAAIL Lab Publications
Poultry Science · 2024 SCI Q1 · IF 4.2 12+ Citations Prior Institutional Research · Pre-IITMZ (2024)

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

Innocent Nyalala 1 , Jiayu Zhang 2 , Zixuan Chen 1 , Junlong Chen 1 , Kunjie Chen 1
1College of Engineering, Nanjing Agricultural University, Nanjing, China
2Xuzhou University of Technology, Xuzhou, China
Poultry Science (Elsevier), Vol. 103, 104232 (Oct 2024)
Online chicken carcass volume estimation using depth imaging and 3-D reconstruction

Summary: Overhead Kinect v2 depth acquisition and 3D point-cloud reconstruction of moving broiler carcasses achieving R^2 = 0.9988, RMSE = 5.335, and 2.125% average relative error in under 1.6 s per carcass.

Abstract

Non-contact online volume estimation is vital for automated grading and throughput optimization in poultry processing plants. We present a depth-imaging and 3-D geometric reconstruction system evaluated under conveyor speeds, establishing strong volume-to-mass correlation without physical contact or line interruption.

Keywords: 3D reconstructiondepth imagingchicken carcassvolume estimationcomputer visionPoultry Science

Key Findings & Contributions

R² = 0.9988
Volume Estimation R²
Bagged tree regression model
2.125%
Average Rel. Error
RMSE = 5.335 on conveyor
< 1.6 s
Processing Time
Per moving carcass on line
Kinect v2 3D
Depth Point Clouds
Surface integration volume
  1. Captured moving chicken carcasses on a live production conveyor using Kinect v2 depth imaging and real-time 3D point-cloud surface reconstruction.
  2. Extracted geometric surface-integration volume alongside 2D and 3D morphological predictors across multiple regression families.
  3. Achieved R² = 0.9988, RMSE = 5.335, and 2.125% average relative error using Bagged Tree regression in under 1.6 seconds per carcass.

Benchmark Results

Online 3-D Chicken Carcass Volume Estimation Results (Poultry Science 2024)

MetricMeasured PerformanceAcquisition & Compute Context
Best Regression ModelBagged Tree RegressionTrained on 2D + 3D point-cloud features
Coefficient of Det. (R²)0.9988Near-perfect volumetric correspondence
Error BoundsRMSE = 5.335 · ARE = 2.125%Non-contact conveyor line grading
Runtime Latency< 1.6 seconds / carcassReal-time Kinect v2 depth + reconstruction

BibTeX

@article{nyalala2024chicken3d,
  title={Online chicken carcass volume estimation using depth imaging and 3-D reconstruction},
  author={Nyalala, Innocent and Zhang, Jiayu and Chen, Zixuan and Chen, Junlong and Chen, Kunjie},
  journal={Poultry Science},
  pages={104232},
  year={2024},
  doi={10.1016/j.psj.2024.104232},
  publisher={Elsevier}
}
APA Citation

Nyalala, I., Zhang, J., Chen, Z., Chen, J., & Chen, K. (2024). Online chicken carcass volume estimation using depth imaging and 3-D reconstruction. Poultry Science, 104232. https://doi.org/10.1016/j.psj.2024.104232