基于点云的地平面初系数估测

提出了一种基于点云的地平面初系数估测算法。采用八叉树存储包含RGB和空间位置信息点云数据,通过比较相邻两帧树的叶节点来获取运动目标的局部点云。并使用随机采样一致性算法估计,由运动目标局部点云重心所构成的点集上的平面模型参数。同时,采用ISOCLUS算法对场景在法向空间和距离空间上进行两次聚类来分割平面。最终,以运动平面法向信息作为参考因素,从待选平面中标识出地平面点云数据。实验结果显示,该方法可有效地估测出地平面初系数。

Initial Coefficients Estimation of Ground Plane with Point Clouds

An approach is proposed for estimating initial coefficients of the ground plane with point clouds.The octree used first to store the point clouds that contain RGB and location information.The moving object's partial point clouds can be acquired by comparing leaf nodes of two neighboring frames.Then a set consists of centroids extracted from moving object's partial point clouds is utilized to fit a plane through RANSAC.Besides,clustering is performed on normal and distance space successively by ISOCLUS algorithm to segment planes.Finally,the ground plane is chosen from segmented planes according to normal of the plane of the moving object.This method achieves reliable results for estimating initial coefficients of the ground plane through experiment.