A Korean Cattle Weight Prediction Approach Using 3D Segmentation-Based Feature Extraction and Regression Machine Learning from Incomplete 3D Shapes Acquired from Real Farm Environments

• DOI: 10.20944/preprints202311.0350.v1
• Publication Date: 2023-11-08T02:29:41Z
• Authors: Chang Gwon Dang; Seung Soo Lee; Mahboob Alam; Sang Min Lee; Mi Na Park; Ha-Seung Seong; Min Ki Baek; Van Thuan Pham; Jae Gu Lee; Seungkyu Han
• Keywords: Feature (linguistics)

Accurate weight measurement is pivotal for monitoring the growth and well-being of cattle. However, conventional weighing process, which involves physically placing cattle on scales, labor-intensive distressing animals. Hence, development automated prediction techniques assumes critical significance. This study proposes a approach Korean using 3D segmentation-based feature extraction regression machine learning from incomplete shapes acquired real farm environments. In initial phase, we generated mesh data multiple-camera system. Subsequently, deep learning-based segmentation with PointNet network model was employed to segment two dominant parts From these segmented parts, three crucial dimensions were extracted. Finally, implemented five models (CatBoost regression, LightGBM, Polynomial Random Forest XGBoost regression) prediction. To validate our approach, captured 270 in various poses, totaling 1190 poses The best result achieved mean absolute error (MAE) 25.2 kg percent (MAPE) 5.81% random forest model.