In this paper, a unified deep convolutional architecture is proposed to address the problems in the person re-identification task. The proposed method adaptively learns the discriminative deep mid-level features of a person and constructs the correspondence features between an image pair in a data-driven manner. The previous Siamese structure deep learning approaches focus only on pair-wise matching between features. In our method, we consider the latent relationship between mid-level features and propose a network structure to automatically construct the correspondence features from all input features without a pre-defined matching function. The experimental results on three benchmarks VIPeR, CUHK01 and CUHK03 show that our unified approach improves over the previous state-of-the-art methods.