<p>Infrastructure inspection is pivotal in ensuring transportation safety, extending service life, and preventing major accidents. This study addresses the challenges associated with inspecting large-span bridges, particularly high-tower structures, by proposing a novel defect identification method that integrates dual-mode UAV camera information. The proposed method consists of two primary stages. In the global information construction phase, a wide-angle camera captures large-scale scene images, while a synthetic strategy based on deep feature representation focuses on high-tower target areas, generating comprehensive images. The resulting full-scale model of the tower surface provides essential macrostructural information for defect analysis. In the local information analysis phase, a zoom camera is employed to achieve precise focus on smaller regions. A segment-based partitioning strategy is introduced to organically integrate global and local details, establishing a multi-scale information framework that balances macro-level structural assessment with micro-level precision analysis. For localized images, a lightweight detection model, Light-YOLO, based on feature ablation and fusion, is proposed to eliminate redundant information while enhancing multi-dimensional connections between spatial and channel features. By harmonizing the holistic examination of large-scale structures with the meticulous representation of local features, the proposed approach ensures both comprehensive structural assessment and fine-grained defect detection. Validation experiments conducted on real bridges demonstrate that the method provides engineers with efficient and reliable data support, significantly enhancing the accuracy and efficiency of bridge inspections.</p>