In the field of clothing change person re-identification (ReID), a pivotal challenge lies in mitigating the interference of varying clothing and extracting discriminative features. Existing methods focus on extracting non-clothing attributes while overlooking global and biometric cues, leading to the loss of critical discriminative features. To tackle this issue, this paper introduces a novel model, the Semantic Coherence Attention Network (SCANet), which comprises two key components: the Semantic Coherence Coordinator (SCC) module and an enhanced feature extraction network. The SCC module is designed to encourage the model to learn more coherent and discriminative pedestrian identity information. By integrating the Multi-Head Attention Semantic Alignment (MHASA) module, it adeptly captures and reconciles the intricate interdependencies among non-apparel features, contour details, and global characteristics. This refinement enhances semantic coherence and markedly amplifies the discriminative power of the feature representation. Furthermore, we have optimized the feature extraction network to meticulously capture pedestrian detail features, thereby improving the model’s overall recognition capabilities. Extensive experiments conducted on benchmark clothing change ReID datasets, namely PRCC and LTCC, demonstrate the remarkable superiority of our approach over existing methods, which validates the effectiveness and superiority of our proposed SCANet model.