【摘要】 水工隧洞等基础设施工程的传统巡检存在检查不及时、作业风险高、人力需求大、检测周期长和误检错检率高等问题。随着人工智能（AI）的发展，智能化巡检技术在工程检测中得到了广泛的应用。为了提高隧洞等低特征环境下的工程巡检效率和智能化水平，提出结合三维激光雷达点云数据改进的Point YOLO识别算法。该算法通过三维激光雷达测距（LiDAR）点云构建虚拟场景，结合Point YOLO识别算法识别水工隧洞表面裂缝，并基于数字孪生技术，将信息与隧洞设计的建筑信息模型（BIM）进行匹配，智能标注物件、缺陷的位置。在裂缝数据集上的实验结果表明：与YOLOv8算法相比，所提算法的裂缝缺陷识别精确度提升了4.7个百分点。可见，所提算法增强了工程巡检方法的可视化能力，简化了巡检流程，保障了工程质量，为工程的智能化高效巡检提供了算法支撑。更多还原

【Abstract】 Traditional inspections of infrastructure projects such as hydraulic tunnels have problems such as untimely inspections, high operational risks, high labor demand, long inspection cycles, and high false and error detection rates. With the development of Artificial Intelligence(AI), intelligent inspection technology has been widely applied in engineering detection. To improve the efficiency and intelligence of engineering inspection in environments with few features such as tunnels, an improved Point YOLO recognition algorithm incorporating three-dimensional LiDAR(Light Detection And Ranging) point cloud data was proposed. In the algorithm, the virtual scene was constructed by three-dimensional laser ranging point cloud, Point YOLO recognition algorithm was combined to recognize cracks on the surface of hydraulic tunnels, and with the help of digital twin technology, the information was matched with Building Information Modeling(BIM) of the tunnel design to mark the locations of objects and defects intelligently. Experimental results on a crack dataset show that the proposed algorithm improves the accuracy of crack defect recognition by 4. 7 percentage points compared with Point YOLOv8 algorithm. It can be seen that the proposed algorithm improves the visualization ability of engineering inspection methods, simplifies the inspection process, and ensures the project quality, which provides an algorithm support of intelligent and efficient engineering inspection.更多还原