基于无人机倾斜摄影和三维激光扫描的桥梁数字化建模方法

A Bridge Digital Modeling Method Based on UAV Tilt Photography and 3D Laser Scanning

为开展无人机倾斜点云与车载激光点云的数据融合,实现融合数据的优势互补,弥补倾斜摄影技术观测视角上的不足,完成桥梁的高精度数字化还原,针对无人机倾斜摄影测量技术无法全面获取完整地物三维信息,导致建模中存在桥梁三维模型局部纹理扭曲、空洞的问题,提出了一种利用无人机倾斜摄影技术与三维激光扫描技术进行多源数据融合建模的方法。首先用倾斜实景三维重建技术将无人机获取的桥梁倾斜实景三角网数据转换为倾斜密集点云。然后针对两种数据分布特点,基于同名平面几何特征的点云配准方法,对待匹配点云中的公共面进行初步提取,并利用多分区最小二乘拟合算法进行去噪。接着对去噪的平面点云使用RANSAC算法进行平面拟合,使用四元数坐标转换模型计算旋转矩阵,建立了间接平差误差方程计算融合参数,将两种点云进行了高精度融合。最后利用全视角覆盖的融合点云数据进行了精细三角网重建,通过纹理自动映射得到融合后的桥梁精细化模型。试验结果表明:影响倾斜实景模型精度的因素主要是数据融合误差和建模误差;基于多源数据融合的建模方法既能保证三维模型的重建效率和精度,又能修正无人机倾斜摄影建模存在的桥梁底部纹理,为桥梁数字化还原提供了可行的方法。

In order to carry out the data fusion of UAV tilt point cloud and vehicle laser point cloud,realize the complementary advantages of fusion data,make up for the deficiencies in the observation angle of tilt photography technology,and complete the high-precision digital restoration of the bridge,in view of the fact that the UAV oblique photogrammetry technology cannot comprehensively obtain the complete 3 D information of ground objects,resulting in the problem of local texture distortion and hollowness in the 3 D bridge model in the modeling,a multi-source data fusion modeling method based on UAV tilt photography technology and 3 D laser scanning technology is proposed.First,the bridge inclined real scene triangulation network data obtained by UAV are converted inclined dense point cloud by using the inclined real scene 3 D reconstruction technology.Then,according to the distribution characteristics of the 2 kinds of data,based on the point cloud registration method with the same name plane geometric features,the common surface in the point cloud to be matched is initially extracted,and the denoising is conducted by using the multi-partition least squares fitting algorithm.Afterwards,the denoised plane point cloud is fitted with the RANSAC algorithm,the rotation matrix is calculated by using the quaternion coordinate transformation model,the error equation of indirect adjustment is established and the fusion parameters are calculated,and the 2 kinds of point clouds are fused with high precision.Finally,the fine triangulation network is reconstructed by using the full view angle covered fused point cloud data,the refined bridge model after fusion is obtained by texture automatic mapping.The experimental result shows that(1) the main factors affecting the accuracy of the tilted reality model are the data fusion error and modeling error;(2) the modeling method based on multi-source data fusion not only can ensure the reconstruction efficiency and accuracy of the 3 D model,but also can correct the bridge bottom texture in UAV tilt photography modeling,which provides a feasible method for bridge digital restoration.