Authors

Abstract

The quality detection of eggs based on deep learning faced many problems, such as similar feature colors and low computational efficiency, which resulted in an increased probability of false detection or missed detection. To effectively solve these problems, this paper proposed an egg quality detection method based on YOLOv8n, which integrated the ContextGuideFusionModule, EfficientHead, and SIOU loss functions by improving the backbone network. The recognition rate from the field test was 88.4%, indicating that the algorithm could meet the real-time monitoring requirements, effectively identify the quality status of eggs, and provide support for intelligent poultry house management.

Abstract in Chinese

为了快速、准确地检测养殖场中鸡蛋的品质，本文提出了一种改进YOLOv8n网络的鸡蛋品质检测模型HCES-YOLO。通过用HGNetV2替换骨[ Zhimin TONG, Prof. Ph.D. Eng.;Shengzhang LI,M.S. Stud. Eng.;Shichuan MIAO,M.S. Stud. Eng.; Tianzhe XU,M.S. Stud. Eng.; Yu ZHOU,M.S. Stud. Eng.; Changhao LI,M.S. Stud. Eng.]干网络，并集成ContextGuideFusionModule和Efficient Head，我们提高了特征提取能力和计算效率。同时，SIOU损失函数进一步改善了小目标检测的性能。实验结果表明，该模型的平均精度（mAP）达到88.6%，比原始网络提高了3.2%，同时浮点运算减少了3.5G，模型参数减少了1.08M。