The invention relates to an enhanced optimization cross-domain pedestrian re-identification method, and belongs to the technical field of computer vision. The method comprises the steps of auxiliary data and target data set construction, reliable training data set and enhanced training data set division, auxiliary data and reliable training data classification loss, reliable training data dynamic neighbor sample exploration and enhanced training data dynamic neighbor sample mining. According to the method, out-of-cluster difficult sample information and in-cluster sample features are fully utilized, potential information of the target domain is mined, and alternate training with auxiliary data is carried out, so that the matching performance of the model in the target domain is improved.