基于改进距离阈值约束的ICP三维配准方法

摘要

针对具有一定公共部分三维数据的配准问题，提出了一种改进的距离阈值约束迭代最近临点（ICP，iterative closest point）算法。该算法通过主成分分析（PCA，principal component analysis）获取参考和输入三维数据的特征向量进行初始变换模型参数设定；在迭代过程中平均残差判断，先后采用距离比率和自适应距离阈值提取公共部分并建立正确的匹配关系，完成三维配准。实验结果表明提出算法配准精度高，迭代收敛性能好，适合于具有部分重叠三维数据的配准。%For the registration of 3D data with certain overlap region, an improved distance threshold constrained ICP（ Iterative Closest Point）algorithm is proposed. Fist, PCA （Principal Component Analysis） is utilized to estimate the initial transformation model parameters. Second, during every iteration, adaptive distance threshold method followed by distance ratio method is utilized to extract the common part for correct corresponding relationships. Then the registration is done. Experiment results show that registration accuracy of our algorithm is high and it is easy to converge, so it is suitable for registration of 3D data with partly overlapped regions.