Robust registration of aerial and close‐range photogrammetric point clouds using visual context features and scale consistency

Point cloud registration is of great significance to the reconstruction high-precision 3D city models. There are some challenges when aligning aerial and close-range photogrammetric point clouds, such as huge view differences caused by different sights sensors, massive noisy points due error in dense matching, scale uncertainty since there no control for absolute orientation. To achieve complementary advantages this paper, a robust cross-source clouds method proposed using image visual context features consistency. First, cross-view matching based on obtain corresponding points. Second, overcome differences, an outlier filtering designed Finally, dual quaternions model considering factor introduced solve spatial transformation model. analyze feasibility qualitatively quantitatively, experiments conducted public scene dataset Dortmund, Germany, Zhengzhou City, China. Three kinds including ground registrations both Dortmund Zhengzhou. The chamfer distances three sets 4.48 m, 5.97 m 4.78 respectively. ablation study shows that improve accuracy at least 14% 25%, demonstrate accomplishes large-scale scenes accurately efficiently, providing solid foundation subsequent fine reconstruction.