The MCAA-YOLO + XPBI integrated model provides a hybrid intelligent measurement method for predicting the body weight of Hu sheep

Accurate measurement of sheep body weight is essential for optimizing feeding management and maximizing economic benefits in modern animal husbandry. Conventional manual methods are inefficient and prone to external influences, leading to unstable measurements that don't meet the needs of intelligent livestock farming. To address this, researchers propose an automated method that estimates body weight from body-size features using an improved YOLOv10 model combined with an XPBI model.

The Multi-Component Attention Aggregation (MCAA) module is integrated into YOLOv10, combining GAM, CBAM, CoordAtt, and ECA to improve feature representation. This results in the MCAA-YOLOv10 model, which achieves high Precision (0.995), Recall (0.986), mAP50 (0.991), and mAP50-95 (0.987) for accurate and stable body-size data acquisition.

Using these features, an integrated prediction model called the XGBoost-PCA-Bagging Integrator (XPBI) is constructed. It employs principal component analysis and guided bagging to enhance dimensionality reduction and parameter selection. The XPBI model demonstrates excellent performance in weight prediction, with a mean absolute error (MAE) of 1.121, root mean square error (RMSE) of 1.490, mean absolute percentage error (MAPE) of 4.47%, and R2 score of 0.918. Inference requires only 0.106 ms per sample, outperforming other comparison models.

This approach confirms the feasibility and effectiveness of predicting Hu sheep body weight from body-size parameters, providing a reliable method for the digital and intelligent transformation of animal husbandry. It enhances the scientific level of livestock management and offers significant social and economic value. The research was supported by various grants and conducted by a team from Tianjin Agricultural University, Qingyang Animal Disease Prevention and Control Center, and others.