Vehicle Re-identification (ReID) aims to accurately identify the same vehicle across different camera views in real-world scenarios, serving as a critical task in intelligent transportation systems. The main challenge arises from significant intra-class variations and inter-class similarities, primarily caused by viewpoint changes, which substantially degrade recognition performance. Although existing feature learning methods have achieved notable progress, they often struggle to capture rich semantic cues and adapt effectively to diverse viewpoints. To address these issues, this paper proposes a novel feature fusion framework for vehicle ReID. Specifically, a Gate-Enhanced Attention Feature Fusion (GAFF) module is introduced to strengthen the discriminative capacity of global features. Additionally, a Perspective-Aware Synergistic Attention (PASA) mechanism is designed to enhance robustness to viewpoint variations. The effectiveness of our proposed approach is validated through extensive experiments on the VeRi-776 and VehicleID benchmarks.