摘要
针对道路病害巡检中存在的检测效率低、定位精度不足以及目标识别重复等问题,本研究提出一种基于改进YOLOv8(You Only Look Once version 8,计算机视觉框架)的无人机道路病害巡检及多特征密度聚类去重方法。构建完整的无人机智能巡检系统,实现从任务规划到数据采集的全流程自动化管理;通过改进YOLOv8的道路病害识别模型,实现典型道路病害的检测与精确定位,有效提升检测精度;结合病害的空间坐标信息、图像特征相似性以及病害种类,提出多特征融合的改进密度聚类算法,解决检测目标的重复问题,显著抑制冗余检测与降低误检,为道路智能养护提供可靠的技术支撑。
To address the challenges of low detection efficiency,insufficient localization accuracy,and duplicate target recognition in road-distress inspection,this study proposes a UAV-based road distress inspection framework that integrates an improved YOLOv8 detector with a multi-feature density-clustering de-duplication method.We build a complete intelligent UAV inspection system to enable end-to-end automation from mission planning to data acquisition.By enhancing the YOLOv8 road-distress recognition model,the system detects typical distress types and achieves precise localization,thereby improving detection accuracy.Furthermore,by fusing spatial coordinates,visual feature similarity,and distress category,we design a multi-feature density clustering algorithm to eliminate duplicate detections,markedly suppress redundancy and reduce false detections,providing reliable technical support for intelligent road maintenance.
作者
顾振东
王琳
葛恒志
卢婕
GU Zhendong;WANG Lin;GE Hengzhi;LU Jie(Zhongdian Hongxin Information Technology Co.,Ltd.)
出处
《江苏通信》
2025年第5期74-81,共8页
Jiangsu Communications
关键词
无人机
道路病害
YOLOv8
密度聚类算法
识别精度
病害去重
unmanned aerial vehicle(UAV)
road distress
YOLOv8
density clustering algorithm
recognition accuracy
distress deduplication
