Objectives Due to significant differences in scanning ranges between different sensors and the limited overlap areas, registration of airborne and ground-based point clouds faces challenges in locating ground-based point clouds within large airborne scanning areas and in feature extraction in overlapping regions. Focusing on this problem, an airborne-ground point cloud registration method is proposed, integrating smartphone positioning and orientation data.

Methods The method combines smartphone positioning and orientation information for coarse registration, and utilizes elevation feature matching under a polar coordinate grid for fine registration. Smartphone positioning and orientation data are used to set the virtual scanning center in the airborne coordinate system and then a polar coordinate grid is constructed for feature construction. The highest visible point in each subgrid is selected to establish a polar elevation feature map for the ground-based point cloud. Based on the subgrid and ring characteristics of the polar elevation feature map, a fitness function is constructed. A genetic algorithm-based method is introduced to search for the optimal virtual scanning center. By performing genetic operations on the airborne map set and identifying the highest fitness value, fine registration of the ground-based point cloud in the airborne coordinate system is achieved.

Results To verify the feasibility of the proposed method, 9 station-based point clouds and 4 handheld-based point clouds are registered with airborne laser point clouds. The results demonstrate that, despite differences in point cloud distribution caused by varying acquisition methods, the fitness variation curves for station-based and handheld point clouds show slight differences, but convergence is achieved within 300 iterations of the genetic algorithm. The final coarse registration accuracy difference does not exceed 0.03 m. Compared to existing algorithms, the registration time is reduced by approximately 60%, with an average registration time of 3.48 min and optimal accuracy reaching 0.016 8 m. Additionally, smartphone positioning and orientation accuracy were verified, showing that when positioning accuracy is within 50 m, the registration success rate exceeds 90%, and within 20 m, the success rate reaches over 98%.

Conclusions Assistance from smartphone positioning and orientation data enables rapid initial registration of ground-based point clouds within large airborne scanning areas. The polar coordinate elevation feature map provides multiple features for low-overlap areas, facilitating the construction of a fitness function and enhancing algorithm robustness. This method can be applied to cross-dataset transfer in urban scenarios.