<p>With the rapid development of computer vision technology, pedestrian re-identification technology has also made significant progress. However, traditional supervised pedestrian re-identification methods rely on a large amount of labeled data, which will consume enormous human and material resources. In this situation, unsupervised cross-domain pedestrian re-identification technology has attracted more and more attention and interest. However, current unsupervised cross-domain pedestrian re-identification technology has a low ability to extract discriminative features, and the noise generated during the clustering process reduces the performance of pedestrian re-identification. To address the above issues, this paper proposes an unsupervised cross-domain pedestrian re-identification method based on squeeze-excitation attention and latent feature mining, aiming to improve the performance of unsupervised cross-domain pedestrian re-identification. Firstly, the squeeze-excitation attention mechanism is utilized to obtain discriminative features of pedestrians. Secondly, the latent feature mining module is utilized to further explore the latent information in discriminative features. Finally, the proposed pseudo label denoising module is utilized to refine clustering results and generate more accurate pseudo labels through residual connections. A large number of experimental results demonstrate the effectiveness of the unsupervised cross-domain pedestrian re-identification method proposed in this paper.</p>