传统的基于二维图像的混凝土结构裂缝量化方法只能获得局部平面裂缝检测结果,很难确定裂缝的全局位置和三维几何信息.现有的基于激光和深度相机的三维采集设备价格昂贵且难以操作.针对这些问题,提出一种基于三维重建的混凝土结构裂缝量化方法,对裂缝边缘的三维位置进行定位.首先,使用改进的SFM(Structure from Motion)算法对多视角图像进行三维场景重建;其次,使用U-Net++模型对二维图像进行裂缝分割识别,进而获得裂缝坐标;最后,将裂缝从二维图像投影到带有裂缝的三维混凝土表面,得到裂缝边缘的三维坐标.针对混凝土墙体进行实验,结果表明,所提出的方法可准确识别混凝土结构裂缝,实现裂缝集成与定位.
Crack Quantification Method of Concrete Structure Based on Three-Dimensional Reconstruction
Traditional two-dimensional image-based crack quantification methods for concrete structures can only yield localized planar crack detection results, but determining the global location and three-dimensional geometric information of the cracks proves challenging. Furthermore, existing three-dimensional acquisition devices, which rely on laser and depth cameras, are costly and complex to operate. To address these challenges, this paper introduces a three-dimensional reconstruction-based crack quantification method for concrete structures that localizes the 3D positions of crack edges. Initially, three-dimensional scene reconstruction is executed using the enhanced Structure from Motion (SFM) algorithm on multi-view images; subsequently, crack segmentation and identification are conducted on the two-dimensional images using U-Net++, which in turn yields the coordinates of the cracks. Ultimately, the cracks are projected from the two-dimensional images onto the three-dimensional concrete surface, enabling the acquisition of the crack edges' 3D coordinates. In this study, experiments were performed on concrete walls, and the results demonstrate that the method introduced herein can accurately delineate cracks in concrete structures and achieve integrated crack characterization and localization.
万方数据
维普
