Abstract: This paper puts forward a reconstruction algorithm based on the error driver，the successive iterative approximation and large-scale 3D scattered data.This paper re-samples the point cloud data，using normalized CSRBF as interpolation function（RBF），then interpolates to a small amount of points data which has been re-sampled.After that the points that don’t attend interpolation will be put into the implicit function equation to calculate error.The points whose error beyond certain threshold will be re-sampled，then they are put into the original sample point set and restart the interpolation.After multiple iterative operation，the minimum points can be used to do interpolation on the original point cloud model.Experimental results show that this algorithm has higher robustness and higher efficiency.

Key words: point cloud reconstruction, error driver, Compactly Supported Radial Basis Function（CSRBF）, surface interpolation

摘要： 提出了一种基于误差驱动的逐次迭代逼近的大规模3D散乱数据的重建算法。首先对点云数据进行重采样，采用归一化的CSRBF作为插值基函数。其次对重采样后少量的点数据进行插值。再次对未参加插值的点带入隐函数方程，计算误差。对误差超过一定阈值的点进行重采样，加入原采样点集合，重新进行插值。这样多次迭代以后便可以用最少的点来插值原来的点云模型。实验结果表明，该算法具有更高的鲁棒性和更高的效率。

关键词: 点云重建, 误差驱动, 紧支撑径向基函数, 曲面插值