基于多薄板样条的多视角非刚性配准算法

发布时间：2018-05-22 11:26

本文选题：多视角非刚性配准 + 深度数据　；参考：《计算机辅助设计与图形学学报》2017年12期

【摘要】：为解决多视角配准中带有低频非刚性形变的深度数据容易产生累积误差、重叠区域未对齐等问题,提出一种基于多薄板样条的多视角非刚性配准算法.首先通过局部迭代最近点刚性配准算法得到重叠视角深度数据之间的对应点;然后基于多薄板样条的全局优化能量公式为每个视角求解一个薄板样条变换,使所有对应点之间距离的平方和最小;最后将优化后的薄板样条变换应用于每个视角的深度数据.通过在优化模型中引入初始点位置约束,该算法能使配准后的数据尽可能保持初始形状.为了加快求解速度,迭代地求解每个薄板样条变换,并且在优化过程中增量式地引入径向基函数.实验室结果表明,文中算法有较高的精度和效率,能够有效地减少累积误差并且提升重叠区域的对齐效果.
[Abstract]:In order to solve the problem that the depth data with low frequency and non rigid deformation in multi view registration is easy to produce cumulative error and the overlapping area is not aligned, a multi view non rigid registration algorithm based on multi thin strip spline is proposed. First, the corresponding point between the depth data of the overlapping angle is obtained by local iterative nearest point rigid registration algorithm. The global optimization energy formula based on the multi plate spline is used to solve a thin plate spline transformation for each angle, making the sum of the square sum of the distance between all corresponding points minimum. Finally, the optimized thin plate spline transformation is applied to the depth data of each angle. As far as possible, the initial shape is maintained as much as possible. In order to speed up the speed of the solution, each plate spline transformation is solved iteratively, and the radial basis function is introduced incrementally in the optimization process. The laboratory results show that the algorithm has high accuracy and efficiency in this paper, which can effectively reduce the cumulative error and improve the alignment effect of the overlapped region.
【作者单位】：浙江大学CAD&CG国家重点实验室;
【基金】：国家自然科学基金(61472349,61732015)
【分类号】：TP391.41

【相似文献】