In order to adapt to concept of precise animal husbandry

an improved K-SVD algorithm is proposed for individual identification of cattle.On the basis of K-SVD algorithm

selection mode of dictionary atoms in sparse encoding phase is changed

making the algorithm more suitable for image classification problems.A cattle face image dataset IMCFR20 containing 20 cattle is constructed and using this dataset to analyze the impact of sparsity and dictionary atomic number on recognition effect.The experimental results show that the improved algorithm has recognition rate of above 90 %

and the recognition effect is greatly improved

which provides a feasible solution for individual recognition of cattle using images.