基于KD-tree优化特征三角形匹配的林地点云配准方法

A registration method for optimizing triplets matching in forest based on KD-tree

针对现有林地配准方法配准效率低的问题,提出一种通过KD-tree加快特征三角形匹配过程的配准方法。首先使用单木位置构建特征三角形,并以边长、顶点协方差矩阵特征值为几何约束条件剔除等腰、等边以及高度共线的三角形,利用KD-tree快速搜索源点云与目标点云之间的同名特征三角形,根据对应同名点计算转换参数,从而实现林地点云配准。实验结果表明,配准效率与RMSE分别平均提升99.2%、83.6%,且算法在含有一定比例单木位置漏检测与误检测时,RMSE均稳定在0.04~0.09 m,证明该方法具有可行性和有效性。

Focusing on the problem of low registration efficiency of existing registration methods in the forest,a registration optimization method based on KD-tree matching triplets is proposed.Firstly,the triplets were constructed using the stem positions.And isosceles,equilateral,and highly co-linear triplets detected by edge lengths and vertex covariance matrix eigenvalues of vertices as geometric constraints were removed.Then the KD-tree was used to quickly search for similar triplets between source point cloud and target point cloud,then calculated the transformation parameters by corresponding point pairs,hence the registration of forest land scanning data was realized.The experimental results show that the running time and RMSE are increased by an average of 99.2%and 83.6%respectively,and when the algorithm contains a certain proportion of stem position leakage detection and false detection,the RMSE is stable at 0.04~0.09 m,this method is proved to be feasible and effective.