Cross-domain person re-identification is an important issue that restricts the application of person re-identification in practice, and the huge difference between the source domain and the target domain is the most critical factor that affects the generalization ability of the model. In response to this problem, we found that semantic components play an important role in cross-domain re-identification. The semantic analysis model is used to extract the features of multiple semantic components of persons, remove the complex and changeable background interference, and improve the cross-domain adaptive ability of the model. At the same time, the use of fine semantic component features to achieve the purpose of component alignment, but also improves the expressive ability of features, thereby improving the generalization ability of the model. Finally, we conducted a large number of cross-domain person re-identification verification experiments between the two person re-identification data sets of Market1501 and DukeMTMC-reID, which proved the effectiveness of our method.