Abstract: In order to promote the development of precision animal husbandry and discuss the longterm span of animal individual recognition, in this paper, the same batch of yak individual image datasets with an interval of 6 months and 12 months are constructed. In the experiment, the PCB+SEResNet50 recognition model with attention mechanism was used to realize shortterm and longterm yak individual recognition, so as to analyze the factors affecting longterm yak individual recognition. The recognition results of this longterm dataset were compared with those of ViT and PGCFL models. The results showed that the mean average precision of the model reached 60.37% and 41.56% on the data set with an interval of 6 months and 12 months. Compared with ViT, it was 1.64% and 5.82% higher, respectively, and compared with PGCFL, it was 12.40% and 11.22% higher, respectively. This study can provide theoretical basis and method guidance for longterm yak individual identification, breeding information and precision management of livestock.

Key words: precision animal husbandry, yak, individual identification, attention mechanism, animal biometrics

摘要： 为推动精准畜牧业的发展及探讨长时间跨度下的动物个体识别，构建间隔6个月和12个月的同一批牦牛个体图像数据集。试验采用引入注意力机制的PCB+SEResNet50识别模型，实现短时和长时牦牛个体识别，从而分析影响长时牦牛个体识别的因素，并在该长时数据集上与ViT和PGCFL模型识别结果进行比较。结果表明:该模型在间隔6个月和12个月的数据集上识别平均精度均值达到60.37%、41.56%。相较于ViT，分别提高1.64%、5.82%；相较于PGCFL，分别提高12.40%、11.22%。该研究可为长时牦牛个体识别、养殖信息化及牲畜精准管理等提供理论依据和方法指导。

关键词: 精准畜牧业, 牦牛, 个体识别, 注意力机制, 动物生物特征