AIGGIII-DO: Adaptive IGGIII Robust Estimation Method Based on Damping Optimization for Promoting the Application of Spherical Targets

Spherical targets have important application value in large-scale terrestrial laser scanning (TLS). Spherical target has rotational invariance, and its center is often used as the characteristic point of point cloud orientation. The fitting precision of the target center severely limits the layout distance and measurement efficiency. In this work, we propose an adaptive IGGIII robust estimation method based on damped optimization (AIGGIII-DO) to improve the fitting precision of a sphere center and expand the use range of the target. This method uses the data characteristics of the residual to update the critical value adaptively and introduces the damping parameter to avoid giving a small weight to the normal observation residual in the first iteration because of the inaccurate initial value. AIGGIII-DO is a "coarse to fine" iterative weighted least-squares method. In the iterative process, the damping parameter is continuously reduced to improve the robustness of the algorithm to address data with high outlier rates. To further optimize the layout scheme of the targets and improve the measurement efficiency, the factors affecting the measurement precision of the spherical target point clouds are analyzed, and a limit measurement distance (LMD) model of the maximum layout distance of the target with respect to the measurement parameters is constructed. The correctness and feasibility of the proposed methods are verified via extensive simulations and real experiments. The results show that AIGGIII-DO is more robust and suitable for practical use than many existing advanced algorithms. AIGGIII-DO can effectively process data containing 80% non-adversarial outlier points, and its operation efficiency is 2~3 orders of magnitude higher than that of RANSAC. The LMD model can estimate the maximum layout distance of the target with a mean error of less than 1 m. Under the same conditions, the LMD model can greatly reduce the number of stations and improve the measurement efficiency of TLS while satisfying millimeter-level measurement precision.