【摘要】 针对顺序序列配准效率较低且会产生误差累计的问题,提出一种三维扫描系统中复杂型面约束的多视角点云配准序列确定方法,通过主成分分析度量点云的局部形貌平坦程度,分割平坦区域点云并提取距离质心最近点作为保留点进行自适应简化;结合平坦区域子集的样点规模及子集中所有样点到拟合平面的偏差均值量化待配准点云的形貌复杂度,并通过形貌复杂度确定配准序列.实验结果表明,该方法能够在多视角点云配准过程中避免累积误差的产生,有效提高多视角点云的整体配准精度.更多还原

【Abstract】 A method for determining registration sequence of multi-view point clouds based on complex surface constraints in three-dimensional scanning system is proposed to solve the problems that there are error accumulation and low efficiency of sequential registration. The local shape smoothness of point clouds is measured by principal component analysis,and the point clouds of smooth areas are segmented and the nearest points from the center of mass are extracted to simplify the point clouds adaptively. The shape complexity of point clouds is quantified by combining the size of the subset of smooth areas and the mean deviation of all the samples from the subset to the fitting plane. Lastly the registration sequence is determined by the shape complexity. The results of the experiments show that this method can avoid cumulative errors in the registration process of multi-view point clouds,and effectively improve the overall registration accuracy.更多还原