39 / 2025-03-31 17:41:42

Automated Estimation of Residual Feed in Layer-Cage Laying Hen Houses: A Deep Learning Approach

layer-cage farming, residual feed estimation, deep learning, automated monitoring, precision livestock farming

2. 畜禽设施养殖工艺与装备系统 > （3）数字化和智慧养殖设施设备

Traditional feed management in layer-cage laying hen houses faces challenges such as labor inefficiency, feed waste, and environmental risks due to manual monitoring. To address these issues, this study proposes an automated method and system for estimating residual feed in layer-cage troughs. A mobile data acquisition system, equipped with adjustable cameras and lighting, captures high-resolution videos of multi-tiered troughs. A deep learning-based feed residue prediction model (YOLOv8n-seg) is trained on annotated datasets to detect cage pillars, trough regions, and residual feed areas. The model processes video frames to localize pillars and segment feed residues, enabling precise calculation of residual proportions via pixel-based analysis. Key innovations include adaptive camera positioning, dual-region segmentation (upper/lower troughs), and a rational pillar screening algorithm to enhance spatial accuracy. Experimental results demonstrate that the system achieves high precision in feed estimation (mAP50 > 90%), significantly improving feed management efficiency while reducing labor costs and waste. By integrating real-time monitoring and data-driven decision support, this approach optimizes feeding strategies, enhances hen welfare, and promotes sustainable poultry farming. This work provides a scalable solution for modern intensive poultry production systems.