适应性距离函数与迭代最近曲面片精细配准

发布时间：2018-03-17 22:07

本文选题：复杂曲面　切入点：数据配准　出处：《浙江大学学报(工学版)》2017年10期 　论文类型：期刊论文

【摘要】：针对复杂曲面物体多视角激光扫描点云数据,提出从深度图像到完整几何模型的配准方法.根据空间点相对位置在刚体变换下的不变特性,利用曲率不变特征和归一化零均值互相关系数构造有效的初始匹配点对数组.基于单位四元数对匹配的特征点对进行坐标变换求解,完成数据粗略配准.探讨改正系数的确定方法与步骤,计算不同改正系数下的均值误差,得到最佳改正系数.运用适应性距离函数和改进迭代最近曲面片精细匹配技术,将不同视角点云在三维空间进行最优化匹配.根据匹配结果计算配准误差,对配准精度和速度进行统计分析.数值试验结果表明,该方法在保证配准精度的前提下能够有效地提高配准效率.
[Abstract]:A registration method from depth image to complete geometric model is proposed for multi-angle laser scanning point cloud data of complex curved surface object. According to the invariance of relative position of space point under rigid body transformation, An effective initial matching point pair array is constructed by using curvature invariant feature and normalized zero mean correlation number. The coordinate transformation of the matching feature point pair based on the unit quaternion pair is carried out. The method and procedure of determining correction coefficient are discussed, the mean error under different correction coefficient is calculated, and the best correction coefficient is obtained. The adaptive distance function and improved iterative nearest surface fine matching technique are used. The point clouds with different angles of view are optimized in 3D space. The registration errors are calculated according to the matching results, and the registration accuracy and velocity are analyzed statistically. The numerical results show that, This method can effectively improve registration efficiency on the premise of ensuring registration accuracy.
【作者单位】：贵州财经大学信息学院;
【基金】：国家自然科学地区基金资助项目(41261094) 贵州省科技厅科学技术基金资助项目(黔科合基础[2016]1020) 贵州省教育厅自然科学拔尖人才基金资助项目(黔教合KY字[2016]069)
【分类号】：TN249;TP391.41

【相似文献】