Missing-Bolt Detection in Highway Corrugated Steel Beam Guardrails Using a Lightweight Improved YOLOv8 Algorithm

Detection of missing bolts in highway corrugated steel beam guardrails is of great significance for ensuring road safety. However, practical inspection still faces several challenges, including small bolt targets, complex guardrail backgrounds, significant illumination variation, partial occlusion, and limited computing resources of vehicle-mounted inspection platforms. To address these problems, this paper proposes a lightweight attention-enhanced YOLOv8 algorithm for missing-bolt detection in highway corrugated steel beam guardrails. In the proposed method, MobileNetV3 is adopted as the backbone network to reduce model complexity and improve inference efficiency. The C2f (cross-stage feature fusion) module is introduced to enhance feature reuse and multi-scale feature representation, while an EMA (efficient multi-scale attention) mechanism is incorporated to strengthen bolt-related features and suppress complex background interference. To verify the engineering applicability of the proposed algorithm, we collected guardrail bolt images using a guardrail inspection robot in laboratory and real highway environments, and the trained model is deployed on a Jetson TX2 embedded platform for field inference testing. Experimental results show that the improved YOLOv8 model achieves a mAP@50 of 98.72%, a Precision of 97.48%, a Recall of 98.68%, and a detection speed of 45 FPS. After embedded deployment and acceleration, the algorithm effectively identifies missing and normal bolts in practical inspection scenarios. The results demonstrate that the proposed method provides an accurate and lightweight solution with practical deployment potential for automated missing-bolt detection in highway corrugated steel beam guardrails, offering technical support for the intelligent maintenance of highway safety facilities.