Robust Algorithm for Fitting Sphere to 3d Point Clouds in Terrestrial Laser Scanning

Sphere target is one of the most important tools in terrestrial laser scanning. Different point clouds obtained form different views can be transformed into the same coordinate system frame by using it, where the determination of sphere centre is a key. Because the traditional estimation method---least-squares method does not have the ability to resist outliers, and results in poor estimated parameter, a new method to determine sphere centre is presented. The method is based on M-estimation, and the estimated parameters are not affected by outliers of point clouds. Experiments using simulation data and real data are conducted. Results show that the proposed method can overcome the effect of outliers, and improve the accuracy of sphere centre coordinates. Compared with least square estimation, it has the advantage of robust.