基于点云高斯曲率场的桥面病害检测筛分方法  

Bridge deck disease detection and screening method based on point cloud Gaussian curvature field

作　　者：唐亮[1] 张翔宇 吴桐 刘一军 李欣昱 TANG Liang;ZHANG Xiangyu;WU Tong;LIU Yijun;LI Xinyu(School of Civil Engineering,Chongqing Jiaotong University,Chongqing 400074,China)

机构地区：[1]重庆交通大学土木工程学院,重庆400074

出　　处：《公路交通科技》2025年第4期94-104,共11页Journal of Highway and Transportation Research and Development

基　　金：国家自然科学基金项目(51708068);重庆英才计划·创新创业示范团队项目(CQYC201903204)。

摘　　要：【目标】传统桥面病害检测主要依赖人工目视检查,耗时费力,易受检测人员经验、主观判断等因素影响,导致检测结果一致性和准确性较难保证。针对于此,本研究提出一种基于高斯曲率场的桥面铺装病害检测方法,以提高检测效率并实现病害的自动识别与分类量化。【方法】首先对三维激光扫描获取的桥面点云数据进行区域分割,提取桥面铺装区域,避免形态边缘对高斯曲率计算的干扰。再利用微分几何方法计算高斯曲率值,构建高斯曲率场,通过分析其异常点分布,对病害进行初步识别与量化。进一步地,结合Hu矩及形态学特征对识别出的异常区域进行二次筛选,以区分真实病害与人工设置物,提高识别精度。将该方法应用于一座混凝土连续梁桥的桥面铺装病害检测分析了其桥面铺装区域的高斯曲率值分布情况和桥梁表观缺陷特征。【结果】识别结果与人工目视检测结果最大相对误差为3.3%,且该方法不受光照条件影响,可代替直接的人工视觉检测。【结论】通过高斯曲率的分布能够表征该桥梁的桥面铺装坑槽、酥松剥落等表面病害,且该方法与传统检测方法对病害程度的评估结果吻合,能够高效量化桥梁结构表面病害,建立结构可视化电子档案,可实施性高,为桥面病害评估提供了一种新的技术途径。[Objective]The traditional bridge deck disease detection mainly relies on manual visual inspection,which is time-consuming,labor-intensive,and highly dependent on the experience and subjective judgment of inspectors.This often results in inconsistencies and reduced accuracy in detection result.To address this issue,the study proposes a bridge deck pavement disease detection method based on Gaussian curvature field to enhance the detection efficiency,and enable the automatic disease identification and classification.[Method]First,the bridge deck point cloud data,obtained with 3D laser scanning,was divided into regions.The bridge deck pavement regions were extracted to avoid the interference of structural edge on Gaussian curvature calculation.Then,by using differential geometry methods,the Gaussian curvature values were calculated,and the curvature field was constructed.The distribution of abnormal curvature points was analyzed to preliminarily identify and quantify the disease.Furthermore,Hu moments and morphological features were employed for secondary screening of identified abnormal regions to distinguish real diseases from artificial settings,and improve the identification accuracy.The proposed method was applied to disease detection of a concrete continuous girder bridge deck.The Gaussian curvature value distribution and bridge pavement defect characteristics in deck pavement regions were analyzed.[Result]The identification result indicates the maximum relative error of 3.3%compared with the manual visual inspection.The proposed method is not affected by lighting conditions.It is a viable alternative to direct manual inspection.[Conclusion]The Gaussian curvature distribution effectively characterizes the bridge deck disease,e.g.,potholes and spalling.The proposed method is consistent with the traditional detection method in bridge deck pavement disease quantification and evaluation.It can efficiently quantify the bridge structural surface disease,establish the visualized electronic file of structure,and has

关 键 词：桥梁工程 桥面铺装病害 三维激光点云 高斯曲率 形态学特征向量 

分 类 号：U445.7[建筑科学—桥梁与隧道工程]