A non-contact rabbit temperature detection model based on improved YOLO11

Le résumé fourni par la source

Rabbit farming is a lucrative industry with strong growth potential, yet measuring rabbits' body temperature remains challenging due to stress responses and the labor-intensive nature of manual methods. This study introduces a non-contact temperature measurement approach using an enhanced YOLO11 deep learning model to detect the region of interest (ROI). The model integrates a channel attention mechanism (CBAM) and a bidirectional information fusion network (BIFN) to improve feature representation and multi-scale feature fusion, while depthwise separable convolution (DWConv) reduces model complexity by 42.6%. Compared to the baseline YOLO11s model, this method boosts precision by 0.92%, recall by 1.76%, F1-score by 1.35%, and mAP50 by 1.67%. To account for environmental influences, key parameters like ambient temperature and CO2 concentration are incorporated to refine temperature measurements. Among various machine learning algorithms tested, the random forest model excelled in predicting body temperature, achieving a coefficient of determination (R²) of 0.8448 and a mean absolute error (MAE) of 0.1061°C. This highlights its ability to capture the complex nonlinear relationship between body temperature and environmental factors, making it ideal for accurate temperature detection in rabbits. This research demonstrates the transformative potential of deep learning and machine learning in advancing precision livestock farming.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.