摘要
高速公路波形梁钢护栏螺栓缺失检测对保障道路安全具有重要意义.然而,在实际检测过程中仍面临螺栓目标尺寸小、护栏背景复杂、光照变化明显、局部遮挡以及车载检测平台计算资源受限等挑战.针对上述问题,本文提出一种面向波形梁钢护栏螺栓缺失检测的轻量化注意力增强YOLOv8算法.算法采用MobileNetV3作为骨干网络,以降低模型复杂度并提高推理效率;引入C2f模块以增强特征复用和多尺度特征表达能力;同时融合高效多尺度注意力机制(EMA),以强化螺栓相关特征并抑制复杂背景干扰.为验证所提算法的工程适用性,本文利用护栏检测机器人在实验室和真实高速公路环境中采集护栏螺栓图像,并将训练后的模型部署至Jetson TX2嵌入式平台进行现场推理测试.实验结果表明,改进YOLOv8模型的mAP@50达到98.72%,准确率为97.48%,召回率为98.68%,检测速度为45 FPS.经嵌入式部署与加速后,该算法能够在实际检测场景中有效识别缺失螺栓和正常螺栓,为高速公路波形梁钢护栏螺栓缺失自动检测提供了一种准确、轻量且具备实际部署潜力的解决方案.
Detection of missing bolts in highway corrugated steel beam guardrails is of great significance for ensuring road safety.How-ever,practical inspection still faces several challenges,including small bolt targets,complex guardrail backgrounds,significant illumina-tion 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 guard-rails.In the proposed method,MobileNetV3 is adopted as the backbone network to reduce model complexity and improve inference effi-ciency.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 inter-ference.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 infer-ence testing.Experimental results show that the improved YOLOv8 model achieves a mAP@50 of 98.72%,a Precision of 97.48%,a Re-call 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.