A Method for Obtaining 3D Point Cloud Data by Combining 2D Image Segmentation and Depth Information of Pigs.

Area of Science:

• Agricultural Engineering
• Computer Vision
• Animal Science

Background:

• Precision livestock farming requires accurate monitoring of individual animals.
• Automatic detection and segmentation of pigs are crucial for efficient farm management.
• Existing methods may lack the precision needed for fine-grained analysis and 3D data acquisition.

Purpose of the Study:

• To develop an advanced method for automatic pig detection and segmentation using RGB-D data.
• To enhance feature extraction for improved accuracy in 2D and 3D pig analysis.
• To facilitate simpler and more efficient acquisition of 3D point cloud data for pigs.

Main Methods:

• Integration of the enhanced YOLOv5s model with the Res2Net bottleneck structure.
• Utilizing RGB-D data for combined 2D image analysis and 3D point cloud generation.
• Development and utilization of custom datasets and the Edinburgh pig behaviour dataset for validation.

Main Results:

• The improved YOLOv5s_Res2Net model achieved high mean Average Precision (mAP) scores for detection and segmentation.
• Achieved mAP@0.5:0.95 of 89.6% and 84.8% on the custom dataset.
• Achieved mAP@0.5:0.95 of 93.4% and 89.4% on the Edinburgh pig behaviour dataset.

Conclusions:

• The proposed method significantly enhances the precision of pig detection and segmentation.
• The approach offers a streamlined process for obtaining 3D pig data.
• This technology supports improved pig management, welfare assessment, and weight estimation.