37 / 2025-03-31 17:50:59

A deep-learning based counting and weighing method for eggs in egg collectors based on deep learning

image segmentation，,Egg weight prediction,Egg counting,deep learning

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

This paper develops a method and apparatus for egg counting and precise weight measurement in layer houses. The method involves acquiring video data containing egg images on a conveyor belt. The egg images are annotated based on contour information to create labeled datasets, which are then used to compose a deep learning segmentation dataset. A segmentation model for egg images is trained and integrated with the Bytetrack object tracking model to achieve egg tracking and counting. Additionally, an egg weight dataset is constructed and utilized to train a regression model. The output of the egg tracking model is linked with the input of the weight prediction model to obtain both the total number of eggs in the video and the individual weight of each egg. The proposed method can effectively track multiple eggs in real time while accomplishing counting and weighing tasks, achieving a counting accuracy of over 99.0% and a weighing accuracy of over 97.8%. Our research can significantly enhance the integration of livestock farming practices.