Body Weight Estimation for Pigs Based on 3D Hybrid Filter and Convolutional Neural Network

Area of Science:

• Agricultural Engineering
• Computer Vision
• Animal Science

Background:

• Accurate pig weight measurement is vital for optimizing growth, health, and feeding strategies.
• Conventional manual weighing is inefficient, time-consuming, and stressful for pigs.
• Existing 2D and 3D methods for weight estimation have limitations in equipment complexity and data processing.

Purpose of the Study:

• To develop a precise and efficient pig weight estimation model using a hybrid 3D point cloud denoising approach.
• To mitigate weight estimation bias and enhance feature extraction accuracy.
• To provide a non-invasive and accurate method for real-time pig weight monitoring.

Main Methods:

• A hybrid 3D point cloud denoising technique combining statistical filtering and DBSCAN clustering.
• Utilizing the convex hull technique to isolate the pig's back for refined data.
• Employing voxel down-sampling for improved real-time processing efficiency.
• Integrating pig back parameters with a Convolutional Neural Network (CNN) for weight prediction.

Main Results:

• The proposed model achieved a Mean Absolute Error (MAE) of 12.45 kg, Mean Absolute Percent Error (MAPE) of 5.36%, and Root Mean Square Error (RMSE) of 12.91 kg.
• The method demonstrated simplified equipment configuration and reduced data processing complexity compared to existing techniques.
• Accuracy of weight estimation was maintained without compromising efficiency.

Conclusions:

• The hybrid 3D point cloud denoising approach offers an effective solution for precise pig weight estimation.
• This method supports scientific feeding management, reduces labor costs, and enhances pig welfare.
• The proposed technique provides a valuable tool for precision livestock farming and intelligent animal husbandry.