FR-ICP算法在倾斜摄影测量点云配准中的应用

Application of the FR-ICP Algorithm in Point Cloud Registration of Tilt Photogrammetry Technology

倾斜摄影测量技术在大场景三维重建应用中,需要通过多片点云数据的刚性配准来实现数据融合。传统迭代最近点(Iterative Closest Point,ICP)算法在实际应用中存在收敛速度慢、易受异常值影响、精度低等具体问题。文章将一种快速且鲁棒的ICP算法(FR-ICP)引入航空倾斜摄影测量应用领域,通过引入Welsch函数描述配准误差,提高缺失点云数据的配准精度;通过李代数形式表示变换矩阵,将优化求解过程转换为固定点迭代加速,大大提高了算法的收敛速度。在小雁塔真实测绘三维点云数据集上验证了算法有效性,实验结果表明FR-ICP算法可解决航空倾斜摄影在古建筑数字化保护应用中多源、大尺度海量点云数据融合处理问题,提高基于倾斜摄影测量重建三维点云的数据精度。

In the application of tilted photogrammetry technique in the large-scale 3D reconstruction, it is necessary to conduct the rigid registration of multiple point cloud data to realize the data fusion. The traditional Interative Closest Point(ICP) algorithm has some problems in practical application, such as slow convergence speed, easy to be affected by outliers and low accuracy. FR-ICP algorithm introduced Welsch function to describe the registration error, which can significantly improve the registration accuracy of missing point cloud data. Meanwhile, the FR-ICP algorithm represents the transformation matrix by Lie algebra form, and the convergence speed of the algorithm is greatly improved based on fixed point iteration acceleration.Experimental results show that the FR-ICP algorithm can improve the data accuracy of 3D point cloud reconstruction based on tilt photogrammetry. By utilizing the FR-ICP algorithm, we can solve the problem of multi-source and large-scale massive point cloud data fusion processing in practical applications, especially for the digital protection of Chinese ancient architectures.