Considering the characteristics of cross-camera retrieval in person re-identification tasks, manual annotation of large datasets is expensive and time-consuming. However, current deep learning-based person re-identification methods still rely on large-scale datasets to prevent network overfitting. An effective way to expand datasets is to employ data augmentation techniques, which are primarily divided into two main categories: generation-based data augmentation and transformation-based data augmentation. The former mainly utilizes generative models to produce new data, but the complexity of their training is often disproportionate to the performance achieved, frequently resulting in diminishing returns.