利用超体素的车载激光点云杆状目标的提取

Extracting Pole-Like Objects in Mobile Laser Scanning Data with Supervoxels

提出了一种利用超体素来提取杆状目标的方法。首先将车载激光点云中的非地面点进行体素化,并在此基础上构建超体素;然后对超体素内的点云进行类型判别,点云大致可分为线型、面型、体型3种类型,杆状目标近似垂直于地面,其点云类型一般为线型;由于杆状目标在水平方向上具有一定的独立性,可以通过一种自适应半径的模型来检测主方向近似平行于Z轴的线型超体素是否属于杆状目标;最后依据各类杆状目标的高度和形状的差异,对杆状目标进行细分类。利用该方法对某城市街区进行杆状目标提取,杆状目标的检测率为93.5%,证明了此方法的可行性。

A method for extracting pole-like objects with supervoxels is proposed.Firstly,the non-ground points from the vehicle-borne laser point clouds are voxelized.On this basis,supervoxels are constructed.Then,the point clouds in supervoxels are classified into three types:linear,planar and spherical.The pole-like object is perpendicular to the ground,so it is generally linear.Since the pole-like objects have a certain degree of independence in the horizontal direction,it is possible to detect whether the linear supervoxels whose main direction approximately parallels to the Zaxis belong to the pole-like objects by an adaptive radius model.Finally,according to the differences of various types of polelike objects in height and shape,the pole-like objects are classified in detail.The method is used to extract pole-like objects of a city block,and the detection rate of pole-like objects is 93.5%,which proves its feasibility.