基于三维点云重建的混凝土结构裂缝定位与追踪

Crack localization and tracking in concrete structures based on 3D point cloud reconstruction

为实现混凝土结构裂缝的自动化定位、宽度量化及扩展追踪,提出一种基于三维点云重建的混凝土结构裂缝定位及扩展追踪方法。首先通过无人机搭载高分辨率云台相机获取目标建筑物的图像集,再通过优化数据集与三维重建流程得到准确的建筑结构点云模型,并还原相机空间参数;然后提出视点定位算法,基于还原的相机空间参数求得拍摄裂缝的相机世界坐标,将裂缝图片与相机世界坐标绑定,基于图片索引裂缝的三维坐标,实现裂缝在点云模型中的自动定位;最后提出适用于混凝土结构的点云映射与配准算法,对裂缝宽度的扩展进行量化追踪。通过试验对服役期的大型混凝土建筑结构进行了可行性和精度验证。结果表明:所提出方法的三维模型重建的尺度平均误差小于3%,且可自动化定位结构裂缝的三维坐标,裂缝平均定位时间为38.09μs;通过进一步将整体模型与更新的裂缝点云集配准,可实现裂缝扩展信息(裂缝宽度)的准确追踪,试验相对误差小于8%。

In order to realize automatic crack localization,width quantification,and extension tracking in concrete structure,a methodology for crack localization and tracking in concrete structures based on three-dimensional(3D)point cloud reconstruction was proposed.Firstly,the image set of the target building was obtained by an unmanned aerial vehicle(UAV)equipped with a high-resolution gimbal camera.Image acquisition and 3D reconstruction process were optimized to obtain an accurate point cloud model of the building structure and restore the camera space parameters.Secondly,the viewpoint localization algorithm was proposed to obtain the camera world coordinates of the cracks based on the restored camera space parameters.After binding the crack images to the world coordinates of cameras,the 3D coordinates of the crack were indexed based on the image to realize the automatic localization of the crack in the point cloud model.Finally,point cloud mapping and the registration algorithm for concrete structures was proposed to quantitatively track the propagation of crack width.The feasibility and accuracy of the large concrete building structure in service period were verified by experiments.The results show that the average scale error of 3D model reconstruction by the method is less than 3%,and the 3D coordinates of structure cracks can be automatically located.The average localization time of cracks is 38.09μs,and by further registering the whole model with the updated crack point cloud set,the accurate tracking of crack propagation information(crack width)can be realized,and the test relative error is less than 8%.