Cross-view geo-localization (CVGL) presents significant challenges due to the drastic variations in perspective and scene layout between unmanned aerial vehicle (UAV) and satellite images. Existing methods primarily emphasize global semantic feature extraction, but often overlook fine-grained local regions and struggle to align cross-view features, limiting their ability to capture discriminative target information. To address this issue, we propose a Structural Perception Enhancement (SPE) network for CVGL. Built upon the DINOv2 backbone, the network integrates a Local Region Mining Module (LRMM) for extracting discriminative regional features and enabling accurate cross-view feature alignment. Furthermore, we introduce a Sample Rebalancing Strategy(SRS) to address training instability caused by satellite image scarcity and sample imbalance. Extensive experiments on the University-1652 and SUES-200 datasets show that our method surpasses existing state-of-the-art approaches, with average improvement of 0.44% in R@1 and 0.96% in AP, validating its effectiveness and superiority.