To better construct walnut core germplasm, this paper uses 245 walnut germplasm resources as test materials and uses the genetic variation data of quantitative traits to explore its construction methods. We used random sampling strategy, deviation degree sampling strategy, and site priority sampling strategy, combined with 8 gradients (5%, 10%, 15%, 20%, 25%, 30%, 40% and 50%). The sampling ratio of the core collection was constructed by clustering multiple times. The core collection was compared with the original collection and the reserved collection by ttest. The core collection was evaluated, and the core collection was confirmed by principal component analysis and phenotypic traits. The results showed that: (1) the deviation degree sampling strategy was selected, the overall sampling ratio was 50%, and 131 core germplasm resources of walnut group Guizhou walnut plants were constructed. (2) The principal component analysis of 13 quantitative traits in the core pool of germplasm resources was carried out, and the cumulative contribution rate reached more than 76.48%. The distribution of walnut core germplasm and original germplasm in the principal component diagram was similar. In this way, the genetic structure of the original walnut germplasm is effectively preserved, and the redundancy of the germplasm is effectively avoided. (3) Under the clustering conditions of Euclidean distance combined with the intergroup connection method, the deviation degree sampling strategy is the best method to construct walnut core collection.